arXiv Papers with Code in Computer Science (March 2026)

Abstract:
Obstacle avoidance in unmanned aerial vehicles (UAVs), as a fundamental capability, has gained increasing attention with the growing focus on spatial intelligence. However, current obstacle-avoidance methods mainly depend on limited field-of-view sensors and are ill-suited for UAV scenarios which require full-spatial awareness when the movement direction differs from the UAV's heading. This limitation motivates us to explore omnidirectional obstacle avoidance for panoramic drones with full-view perception. We first study an under explored problem setting in which a UAV must generate collision-free motion in environments with obstacles from arbitrary directions, and then construct a benchmark that consists of three representative flight tasks. Based on such settings, we propose Fly360, a two-stage perception-decision pipeline with a fixed random-yaw training strategy. At the perception stage, panoramic RGB observations are input and converted into depth maps as a robust intermediate representation. For the policy network, it is lightweight and used to output body-frame velocity commands from depth inputs. Extensive simulation and real-world experiments demonstrate that Fly360 achieves stable omnidirectional obstacle avoidance and outperforms forward-view baselines across all tasks. Our model is available at https://zxkai.github.io/fly360/

Paperid: 2, https://arxiv.org/pdf/2603.06572.pdf GitHub

Abstract:
Incremental Few-Shot (IFS) segmentation aims to learn new categories over time from only a few annotations. Although widely studied in 2D, it remains underexplored for 3D point clouds. Existing methods suffer from catastrophic forgetting or fail to learn discriminative prototypes under sparse supervision, and often overlook a key cue: novel categories frequently appear as unlabelled background in base-training scenes. We introduce SCOPE (Scene-COntextualised Prototype Enrichment), a plug-and-play background-guided prototype enrichment framework that integrates with any prototype-based 3D segmentation method. After base training, a class-agnostic segmentation model extracts high-confidence pseudo-instances from background regions to build a prototype pool. When novel classes arrive with few labelled samples, relevant background prototypes are retrieved and fused with few-shot prototypes to form enriched representations without retraining the backbone or adding parameters. Experiments on ScanNet and S3DIS show that SCOPE achieves SOTA performance, improving novel-class IoU by up to 6.98% and 3.61%, and mean IoU by 2.25% and 1.70%, respectively, while maintaining low forgetting. Code is available https://github.com/Surrey-UP-Lab/SCOPE.

Paperid: 3, https://arxiv.org/pdf/2603.06569.pdf GitHub

Abstract:
Vision Language Model (VLM) development has largely relied on scaling model size, which hinders deployment on compute-constrained mobile and edge devices such as smartphones and robots. In this work, we explore the performance limits of compact (e.g., 2B and 8B) VLMs. We challenge the prevailing practice that state-of-the-art VLMs must rely on vision encoders initialized via massive contrastive pretraining (e.g., CLIP/SigLIP). We identify an objective mismatch: contrastive learning, optimized for discrimination, enforces coarse and category-level invariances that suppress fine-grained visual cues needed for dense captioning and complex VLM reasoning. To address this issue, we present Penguin-VL, whose vision encoder is initialized from a text-only LLM. Our experiments reveal that Penguin-Encoder serves as a superior alternative to traditional contrastive pretraining, unlocking a higher degree of visual fidelity and data efficiency for multimodal understanding. Across various image and video benchmarks, Penguin-VL achieves performance comparable to leading VLMs (e.g., Qwen3-VL) in mathematical reasoning and surpasses them in tasks such as document understanding, visual knowledge, and multi-perspective video understanding. Notably, these gains are achieved with a lightweight architecture, demonstrating that improved visual representation rather than model scaling is the primary driver of performance. Our ablations show that Penguin-Encoder consistently outperforms contrastive-pretrained encoders, preserving fine-grained spatial and temporal cues that are critical for dense perception and complex reasoning. This makes it a strong drop-in alternative for compute-efficient VLMs and enables high performance in resource-constrained settings. Code: https://github.com/tencent-ailab/Penguin-VL

Paperid: 4, https://arxiv.org/pdf/2603.06544.pdf GitHub

Abstract:
Next-generation autonomous vehicles (AVs) rely on large volumes of multisource and multimodal ($M^2$) data to support real-time decision-making. In practice, data quality (DQ) varies across sources and modalities due to environmental conditions and sensor limitations, yet AV research has largely prioritized algorithm design over DQ analysis. This work focuses on redundancy as a fundamental but underexplored DQ issue in AV datasets. Using the nuScenes and Argoverse 2 (AV2) datasets, we model and measure redundancy in multisource camera data and multimodal image-LiDAR data, and evaluate how removing redundant labels affects the YOLOv8 object detection task. Experimental results show that selectively removing redundant multisource image object labels from cameras with shared fields of view improves detection. In nuScenes, mAP${50}$ gains from $0.66$ to $0.70$, $0.64$ to $0.67$, and from $0.53$ to $0.55$, on three representative overlap regions, while detection on other overlapping camera pairs remains at the baseline even under stronger pruning. In AV2, $4.1$-$8.6\%$ of labels are removed, and mAP${50}$ stays near the $0.64$ baseline. Multimodal analysis also reveals substantial redundancy between image and LiDAR data. These findings demonstrate that redundancy is a measurable and actionable DQ factor with direct implications for AV performance. This work highlights the role of redundancy as a data quality factor in AV perception and motivates a data-centric perspective for evaluating and improving AV datasets. Code, data, and implementation details are publicly available at: https://github.com/yhZHOU515/RedundancyAD

Paperid: 5, https://arxiv.org/pdf/2603.06543.pdf GitHub GitHub

Abstract:
We introduce SurgFormer, a multiresolution gated transformer for data driven soft tissue simulation on volumetric meshes. High fidelity biomechanical solvers are often too costly for interactive use, so we train SurgFormer on solver generated data to predict nodewise displacement fields at near real time rates. SurgFormer builds a fixed mesh hierarchy and applies repeated multibranch blocks that combine local message passing, coarse global self attention, and pointwise feedforward updates, fused by learned per node, per channel gates to adaptively integrate local and long range information while remaining scalable on large meshes. For cut conditioned simulation, resection information is encoded as a learned cut embedding and provided as an additional input, enabling a unified model for both standard deformation prediction and topology altering cases. We also introduce two surgical simulation datasets generated under a unified protocol with XFEM based supervision: a cholecystectomy resection dataset and an appendectomy manipulation and resection dataset with cut and uncut cases. To our knowledge, this is the first learned volumetric surrogate setting to study XFEM supervised cut conditioned deformation within the same volumetric pipeline as standard deformation prediction. Across diverse baselines, SurgFormer achieves strong accuracy with favorable efficiency, making it a practical backbone for both tasks. {Code, data, and project page: \href{https://mint-vu.github.io/SurgFormer/}{available here}}

Paperid: 6, https://arxiv.org/pdf/2603.06495.pdf GitHub

Abstract:
Activation steering methods enable inference-time control of large language model (LLM) behavior without retraining, but current approaches face a fundamental trade-off: sample-efficient methods suboptimally capture steering signals from labeled examples, while methods that better extract these signals require hundreds to thousands of examples. We introduce COLD-Steer, a training-free framework that steers LLM activations by approximating the representational changes that would result from gradient descent on in-context examples. Our key insight is that the effect of fine-tuning on a small set of examples can be efficiently approximated at inference time without actual parameter updates. We formalize this through two complementary approaches: (i) a unit kernel approximation method that updates the activations directly using gradients with respect to them, normalized across examples, and (ii) a finite-difference approximation requiring only two forward passes regardless of example count. Experiments across a variety of steering tasks and benchmarks demonstrate that COLD-Steer achieves upto 95% steering effectiveness while using 50 times fewer samples compared to the best baseline. COLD-Steer facilitates accommodating diverse perspectives without extensive demonstration data, which we validate through our experiments on pluralistic alignment tasks. Our framework opens new possibilities for adaptive, context-aware model control that can flexibly address varying loss-driven human preferences through principled approximation of learning dynamics rather than specialized training procedures.

Paperid: 7, https://arxiv.org/pdf/2603.06449.pdf GitHub

Abstract:
Autoregressive (AR) language models rely on causal tokenization, but extending this paradigm to vision remains non-trivial. Current visual tokenizers either flatten 2D patches into non-causal sequences or enforce heuristic orderings that misalign with the "next-token prediction" pattern. Recent diffusion autoencoders similarly fall short: conditioning the decoder on all tokens lacks causality, while applying nested dropout mechanism introduces imbalance. To address these challenges, we present CaTok, a 1D causal image tokenizer with a MeanFlow decoder. By selecting tokens over time intervals and binding them to the MeanFlow objective, as illustrated in Fig. 1, CaTok learns causal 1D representations that support both fast one-step generation and high-fidelity multi-step sampling, while naturally capturing diverse visual concepts across token intervals. To further stabilize and accelerate training, we propose a straightforward regularization REPA-A, which aligns encoder features with Vision Foundation Models (VFMs). Experiments demonstrate that CaTok achieves state-of-the-art results on ImageNet reconstruction, reaching 0.75 FID, 22.53 PSNR and 0.674 SSIM with fewer training epochs, and the AR model attains performance comparable to leading approaches.

Paperid: 8, https://arxiv.org/pdf/2603.06424.pdf GitHub

Abstract:
Large language models (LLMs) have recently reshaped Automated Essay Scoring (AES), yet prior studies typically examine individual techniques in isolation, limiting understanding of their relative merits for English as a Second Language (L2) writing. To bridge this gap, we presents a comprehensive comparison of major LLM-based AES paradigms on IELTS Writing Task~2. On this unified benchmark, we evaluate four approaches: (i) encoder-based classification fine-tuning, (ii) zero- and few-shot prompting, (iii) instruction tuning and Retrieval-Augmented Generation (RAG), and (iv) Supervised Fine-Tuning combined with Direct Preference Optimization (DPO) and RAG. Our results reveal clear accuracy-cost-robustness trade-offs across methods, the best configuration, integrating k-SFT and RAG, achieves the strongest overall results with F1-Score 93%. This study offers the first unified empirical comparison of modern LLM-based AES strategies for English L2, promising potential in auto-grading writing tasks. Code is public at https://github.com/MinhNguyenDS/LLM_AES-EnL2

Paperid: 9, https://arxiv.org/pdf/2603.06401.pdf GitHub

Abstract:
The upper 6 GHz (U6G) band with XL-MIMO is a key enabler for sixth-generation wireless systems, yet intelligent radiomap prediction for such systems remains challenging. Existing datasets support only small-scale arrays (up to 8x8) with predominantly isotropic antennas, far from the 1024-element directional arrays envisioned for 6G. Moreover, current methods encode array configurations as scalar parameters, forcing neural networks to extrapolate array-specific radiation patterns, which fails when predicting radiomaps for configurations absent from training data. To jointly address data scarcity and generalization limitations, this paper advances XL-MIMO radiomap prediction from three aspects. To overcome data limitations, we construct the first XL-MIMO radiomap dataset containing 78400 radiomaps across 800 urban scenes, five frequency bands (1.8-6.7 GHz), and nine array configurations up to 32x32 uniform planar arrays with directional elements. To enable systematic evaluation, we establish a comprehensive benchmark framework covering practical scenarios from coverage estimation without field measurements to generalization across unseen configurations and environments. To enable generalization to arbitrary beam configurations without retraining, we propose the beam map, a physics-informed spatial feature that analytically computes array-specific coverage patterns. By decoupling deterministic array radiation from data learned multipath propagation, beam maps shift generalization from neural network extrapolation to physics-based computation. Integrating beam maps into existing architectures reduces mean absolute error by up to 60.0% when generalizing to unseen configurations and up to 50.5% when transferring to unseen environments. The complete dataset and code are publicly available at https://lxj321.github.io/MulticonfigRadiomapDataset/.

Paperid: 10, https://arxiv.org/pdf/2603.06386.pdf GitHub

Abstract:
Scene Graph Generation (SGG) is a task that encodes visual relationships between objects in images as graph structures. SGG shows significant promise as a foundational component for downstream tasks, such as reasoning for embodied agents. To enable real-time applications, SGG must address the trade-off between performance and inference speed. However, current methods tend to focus on one of the following: (1) improving relation prediction accuracy, (2) enhancing object detection accuracy, or (3) reducing latency, without aiming to balance all three objectives simultaneously. To address this limitation, we build on the powerful Real-time Efficiency and Accuracy Compromise for Tradeoffs in Scene Graph Generation (REACT) architecture and propose REACT++, a new state-of-the-art model for real-time SGG. By leveraging efficient feature extraction and subject-to-object cross-attention within the prototype space, REACT++ balances latency and representational power. REACT++ achieves the highest inference speed among existing SGG models, improving relation prediction accuracy without sacrificing object detection performance. Compared to the previous REACT version, REACT++ is 20% faster with a gain of 10% in relation prediction accuracy on average. The code is available at https://github.com/Maelic/SGG-Benchmark.

Paperid: 11, https://arxiv.org/pdf/2603.06365.pdf GitHub

Abstract:
AI-assisted software generation has increased development speed, but it has also amplified a persistent engineering problem: systems that are functionally correct may still be structurally insecure. In practice, prompt-based security review with large language models often suffers from uneven coverage, weak reproducibility, unsupported findings, and the absence of an immutable audit trail. The ESAA architecture addresses a related governance problem in agentic software engineering by separating heuristic agent cognition from deterministic state mutation through append-only events, constrained outputs, and replay-based verification. This paper presents ESAA-Security, a domain-specific specialization of ESAA for agent-assisted security auditing of software repositories, with particular emphasis on AI-generated or AI-modified code. ESAA-Security structures auditing as a governed execution pipeline with four phases reconnaissance, domain audit execution, risk classification, and final reporting and operationalizes the workflow into 26 tasks, 16 security domains, and 95 executable checks. The framework produces structured check results, vulnerability inventories, severity classifications, risk matrices, remediation guidance, executive summaries, and a final markdown/JSON audit report. The central idea is that security review should not be modeled as a free-form conversation with an LLM, but as an evidence-oriented audit process governed by contracts and events. In ESAA-Security, agents emit structured intentions under constrained protocols; the orchestrator validates them, persists accepted outputs to an append-only log, reprojects derived views, and verifies consistency through replay and hashing. The result is a traceable, reproducible, and risk-oriented audit architecture whose final report is auditable by construction.

Paperid: 12, https://arxiv.org/pdf/2603.06331.pdf GitHub

Abstract:
Diffusion-based world models have shown strong potential for unified world simulation, but the iterative denoising remains too costly for interactive use and long-horizon rollouts. While feature caching can accelerate inference without training, we find that policies designed for single-modal diffusion transfer poorly to world models due to two world-model-specific obstacles: \emph{token heterogeneity} from multi-modal coupling and spatial variation, and \emph{non-uniform temporal dynamics} where a small set of hard tokens drives error growth, making uniform skipping either unstable or overly conservative. We propose \textbf{WorldCache}, a caching framework tailored to diffusion world models. We introduce \textit{Curvature-guided Heterogeneous Token Prediction}, which uses a physics-grounded curvature score to estimate token predictability and applies a Hermite-guided damped predictor for chaotic tokens with abrupt direction changes. We also design \textit{Chaotic-prioritized Adaptive Skipping}, which accumulates a curvature-normalized, dimensionless drift signal and recomputes only when bottleneck tokens begin to drift. Experiments on diffusion world models show that WorldCache delivers up to \textbf{3.7$\times$} end-to-end speedups while maintaining \textbf{98\%} rollout quality, demonstrating the vast advantages and practicality of WorldCache in resource-constrained scenarios. Our code is released in https://github.com/FofGofx/WorldCache.

Paperid: 13, https://arxiv.org/pdf/2603.06279.pdf GitHub GitHub

Abstract:
3D semantic occupancy prediction is a cornerstone of robotic perception, yet real-world voxel annotations are inherently corrupted by structural artifacts and dynamic trailing effects. This raises a critical but underexplored question: can autonomous systems safely rely on such unreliable occupancy supervision? To systematically investigate this issue, we establish OccNL, the first benchmark dedicated to 3D occupancy under occupancy-asymmetric and dynamic trailing noise. Our analysis reveals a fundamental domain gap: state-of-the-art 2D label noise learning strategies collapse catastrophically in sparse 3D voxel spaces, exposing a critical vulnerability in existing paradigms. To address this challenge, we propose DPR-Occ, a principled label noise-robust framework that constructs reliable supervision through dual-source partial label reasoning. By synergizing temporal model memory with representation-level structural affinity, DPR-Occ dynamically expands and prunes candidate label sets to preserve true semantics while suppressing noise propagation. Extensive experiments on SemanticKITTI demonstrate that DPR-Occ prevents geometric and semantic collapse under extreme corruption. Notably, even at 90% label noise, our method achieves significant performance gains (up to 2.57% mIoU and 13.91% IoU) over existing label noise learning baselines adapted to the 3D occupancy prediction task. By bridging label noise learning and 3D perception, OccNL and DPR-Occ provide a reliable foundation for safety-critical robotic perception in dynamic environments. The benchmark and source code will be made publicly available at https://github.com/mylwx/OccNL.

Paperid: 14, https://arxiv.org/pdf/2603.06275.pdf GitHub

Abstract:
Diffusion transformer (DiT) architectures show great potential for real-world image super-resolution (Real-ISR). However, their computationally expensive iterative sampling necessitates one-step distillation. Existing one-step distillation methods struggle with Real-ISR on DiT. They suffer from fundamental trajectory mismatch and generate severe grid-like periodic artifacts. To tackle these challenges, we propose StrSR, a novel one-step adversarial distillation framework featuring spectral and trajectory regularization. Specifically, we propose an asymmetric discriminative distillation architecture to bridge the trajectory gap. Additionally, we design a frequency distribution matching strategy to effectively suppress DiT-specific periodic artifacts caused by high-frequency spectral leakage. Extensive experiments demonstrate that StrSR achieves state-of-the-art performance in Real-ISR, across both quantitative metrics and visual perception. The code and models will be released at https://github.com/jkwang28/StrSR .

Paperid: 15, https://arxiv.org/pdf/2603.06254.pdf GitHub GitHub

Abstract:
Generalizing across unknown targets is critical for open-world perception, yet existing 3D Multi-Object Tracking (3D MOT) pipelines remain limited by closed-set assumptions and ``semantic-blind'' heuristics. To address this, we propose Next-step Open-Vocabulary Autoregression (NOVA), an innovative paradigm that shifts 3D tracking from traditional fragmented distance-based matching toward generative spatio-temporal semantic modeling. NOVA reformulates 3D trajectories as structured spatio-temporal semantic sequences, enabling the simultaneous encoding of physical motion continuity and deep linguistic priors. By leveraging the autoregressive capabilities of Large Language Models (LLMs), we transform the tracking task into a principled process of next-step sequence completion. This mechanism allows the model to explicitly utilize the hierarchical structure of language space to resolve fine-grained semantic ambiguities and maintain identity consistency across complex long-range sequences through high-level commonsense reasoning. Extensive experiments on nuScenes, V2X-Seq-SPD, and KITTI demonstrate the superior performance of NOVA. Notably, on the nuScenes dataset, NOVA achieves an AMOTA of 22.41% for Novel categories, yielding a significant 20.21% absolute improvement over the baseline. These gains are realized through a compact 0.5B autoregressive model. Code will be available at https://github.com/xifen523/NOVA.

Paperid: 16, https://arxiv.org/pdf/2603.06243.pdf GitHub

Abstract:
Group relative policy optimization (GRPO) has become a standard post-training paradigm for improving reasoning and preference alignment in large language models (LLMs), and has recently shown strong effectiveness in LLM-based recommender systems. However, extending GRPO-based reasoning pipelines to multimodal sequential recommendation (MSR) with multimodal large language models (MLLMs) faces fundamental obstacles. First, MSR requires jointly encoding visual content for both historical interactions and multiple candidate items, causing visual tokens to dominate the input and making the cost of group-based rollout scale with history length and candidate set size, which renders GRPO-based training prohibitively expensive. Second, existing Chain-of-Thought (CoT) supervision suffers from reward inflation in recommendation scenarios, where higher training rewards do not reliably translate into improved ranking performance and may induce shortcut learning. To address these challenges, we propose MLLMRec-R1, an efficient and stable GRPO-based reasoning framework for multimodal sequential recommendation. MLLMRec-R1 textualizes visual signals offline to eliminate expensive visual tokens while preserving multimodal semantics, and constructs high-quality multimodal CoT supervision through refinement and confidence-aware assessment. Furthermore, a mixed-grained data augmentation strategy selectively injects reliable CoT samples while retaining standard training data, mitigating reward inflation and improving generalization stability. Extensive experiments on three benchmark datasets demonstrate that MLLMRec-R1 consistently outperforms state-of-the-art methods, establishing a practical and effective GRPO-based reasoning pipeline for multimodal sequential recommendation. The code is available at https://github.com/wangyu0627/MLLMRec-R1.

Paperid: 17, https://arxiv.org/pdf/2603.06242.pdf GitHub

Abstract:
Model merging aims to integrate multiple task-adapted models into a unified model that preserves the knowledge of each task. In this paper, we identify that the key to this knowledge retention lies in maintaining the directional consistency of singular spaces between merged multi-task vector and individual task vectors. However, this consistency is frequently compromised by two issues: i) an imbalanced energy distribution within task vectors, where a small fraction of singular values dominate the total energy, leading to the neglect of semantically important but weaker components upon merging, and ii) the geometric inconsistency of task vectors in parameter space, which causes direct merging to distort their underlying directional geometry. To address these challenges, we propose DC-Merge, a method for directional-consistent model merging. It first balances the energy distribution of each task vector by smoothing its singular values, ensuring all knowledge components are adequately represented. These energy-balanced vectors are then projected onto a shared orthogonal subspace to align their directional geometries with minimal reconstruction error. Finally, the aligned vectors are aggregated in the shared orthogonal subspace and projected back to the original parameter space. Extensive experiments on vision and vision-language benchmarks show that DC-Merge consistently achieves state-of-the-art performance in both full fine-tuning and LoRA settings. The implementation code is available at https://github.com/Tobeginwith/DC-Merge.

Paperid: 18, https://arxiv.org/pdf/2603.06231.pdf GitHub

Abstract:
Trajectory prediction is essential for autonomous driving, enabling vehicles to anticipate the motion of surrounding agents to support safe planning. However, most existing predictors assume fixed-length histories and suffer substantial performance degradation when observations are variable or extremely short in real-world settings (e.g., due to occlusion or a limited sensing range). We propose TaPD (Temporal-adaptive Progressive Distillation), a unified plug-and-play framework for observation-adaptive trajectory forecasting under variable history lengths. TaPD comprises two cooperative modules: an Observation-Adaptive Forecaster (OAF) for future prediction and a Temporal Backfilling Module (TBM) for explicit reconstruction of the past. OAF is built on progressive knowledge distillation (PKD), which transfers motion pattern knowledge from long-horizon "teachers" to short-horizon "students" via hierarchical feature regression, enabling short observations to recover richer motion context. We further introduce a cosine-annealed distillation weighting scheme to balance forecasting supervision and feature alignment, improving optimization stability and cross-length consistency. For extremely short histories where implicit alignment is insufficient, TBM backfills missing historical segments conditioned on scene evolution, producing context-rich trajectories that strengthen PKD and thereby improve OAF. We employ a decoupled pretrain-reconstruct-finetune protocol to preserve real-motion priors while adapting to backfilled inputs. Extensive experiments on Argoverse 1 and Argoverse 2 show that TaPD consistently outperforms strong baselines across all observation lengths, delivers especially large gains under very short inputs, and improves other predictors (e.g., HiVT) in a plug-and-play manner. Code will be available at https://github.com/zhouhao94/TaPD.

Paperid: 19, https://arxiv.org/pdf/2603.06217.pdf GitHub

Abstract:
Residential demand response depends on sustained prosumer participation, yet existing coordination is either fully automated, or limited to one-way dispatch signals and price alerts that offer little possibility for informed decision-making. This paper introduces Conversational Demand Response (CDR), a coordination mechanism where aggregators and prosumers interact through bidirectional natural language, enabled through agentic AI. A two-tier multi-agent architecture is developed in which an aggregator agent dispatches flexibility requests and a prosumer Home Energy Management System (HEMS) assesses deliverability and cost-benefit by calling an optimization-based tool. CDR also enables prosumer-initiated upstream communication, where changes in preferences can reach the aggregator directly. Proof-of-concept evaluation shows that interactions complete in under 12 seconds. The architecture illustrates how agentic AI can bridge the aggregator-prosumer coordination gap, providing the scalability of automated DR while preserving the transparency, explainability, and user agency necessary for sustained prosumer participation. All system components, including agent prompts, orchestration logic, and simulation interfaces, are released as open source to enable reproducibility and further development.

Paperid: 20, https://arxiv.org/pdf/2603.06213.pdf GitHub

Abstract:
Multimodal Summarization (MMS) aims to generate concise textual summaries by understanding and integrating information across videos, transcripts, and images. However, existing approaches still suffer from three main challenges: (1) reliance on domain-specific supervision, (2) implicit fusion with weak cross-modal grounding, and (3) flat temporal modeling without event transitions. To address these issues, we introduce **CoE**, a training-free MMS framework that performs structured reasoning through a **Chain-of-Events** guided by a Hierarchical Event Graph (HEG). The HEG encodes textual semantics into an explicit event hierarchy that scaffolds cross-modal grounding and temporal reasoning. Guided by this structure, **CoE** localizes key visual cues, models event evolution and causal transitions, and refines outputs via lightweight style adaptation for domain alignment. Extensive experiments on eight diverse datasets demonstrate that **CoE** consistently outperforms state-of-the-art video CoT baselines, achieving average gains of **+3.04 ROUGE**, **+9.51 CIDEr**, and **+1.88 BERTScore**, highlighting its robustness, interpretability, and cross-domain generalization. Our code is available at https://github.com/youxiaoxing/CoE.

Paperid: 21, https://arxiv.org/pdf/2603.06205.pdf GitHub

Abstract:
Inertial measurement units (IMUs), which provide high-frequency linear acceleration and angular velocity measurements, serve as fundamental sensing modalities in robotic systems. Recent advances in deep neural networks have led to remarkable progress in inertial odometry. However, the heavy reliance on ground truth data during training fundamentally limits scalability and generalization to unseen and diverse environments. We propose KISS-IMU, a novel self-supervised inertial odometry framework that eliminates ground truth dependency by leveraging simple LiDAR-based ICP registration and pose graph optimization as a supervisory signal. Our approach embodies two key principles: keeping the IMU stable through motion-aware balanced training and keeping the IMU strong through uncertainty-driven adaptive weighting during inference. To evaluate performance across diverse motion patterns and scenarios, we conducted comprehensive experiments on various real-world platforms, including quadruped robots. Importantly, we train only the IMU network in a self-supervised manner, with LiDAR serving solely as a lightweight supervisory signal rather than requiring additional learnable processes. This design enables the framework to ensure robustness without relying on joint multi-modal learning or ground truth supervision. The supplementary materials are available at https://sparolab.github.io/research/kiss_imu.

Paperid: 22, https://arxiv.org/pdf/2603.06200.pdf GitHub

Abstract:
Existing image reflection removal methods struggle to handle complex reflections. Accurate language descriptions can help the model understand the image content to remove complex reflections. However, due to blurred and distorted interferences in reflected images, machine-generated language descriptions of the image content are often inaccurate, which harms the performance of language-guided reflection removal. To address this, we propose the Adaptive Language-Aware Network (ALANet) to remove reflections even with inaccurate language inputs. Specifically, ALANet integrates both filtering and optimization strategies. The filtering strategy reduces the negative effects of language while preserving its benefits, whereas the optimization strategy enhances the alignment between language and visual features. ALANet also utilizes language cues to decouple specific layer content from feature maps, improving its ability to handle complex reflections. To evaluate the model's performance under complex reflections and varying levels of language accuracy, we introduce the Complex Reflection and Language Accuracy Variance (CRLAV) dataset. Experimental results demonstrate that ALANet surpasses state-of-the-art methods for image reflection removal. The code and dataset are available at https://github.com/fashyon/ALANet.

Paperid: 23, https://arxiv.org/pdf/2603.06198.pdf GitHub

Abstract:
Retrieval-Augmented Generation (RAG) is a framework in which a Generator, such as a Large Language Model (LLM), produces answers by retrieving documents from an external collection using a Retriever. In practice, Generators must integrate evidence from long contexts, perform multi-step reasoning, interpret tables, and abstain when evidence is missing. However, existing benchmarks for Generators provide limited coverage, with none enabling simultaneous evaluation of multiple capabilities under unified conditions. To bridge the gap between existing evaluations and practical use, we introduce LIT-RAGBench (the Logic, Integration, Table, Reasoning, and Abstention RAG Generator Benchmark), which defines five categories: Integration, Reasoning, Logic, Table, and Abstention, each further divided into practical evaluation aspects. LIT-RAGBench systematically covers patterns combining multiple aspects across categories. By using fictional entities and scenarios, LIT-RAGBench evaluates answers grounded in the provided external documents. The dataset consists of 114 human-constructed Japanese questions and an English version generated by machine translation with human curation. We use LLM-as-a-Judge for scoring and report category-wise and overall accuracy. Across API-based and open-weight models, no model exceeds 90% overall accuracy. By making strengths and weaknesses measurable within each category, LIT-RAGBench serves as a valuable metric for model selection in practical RAG deployments and for building RAG-specialized models. We release LIT-RAGBench, including the dataset and evaluation code, at https://github.com/Koki-Itai/LIT-RAGBench.

Paperid: 24, https://arxiv.org/pdf/2603.06183.pdf GitHub

Abstract:
We introduce CRIMSON, a clinically grounded evaluation framework for chest X-ray report generation that assesses reports based on diagnostic correctness, contextual relevance, and patient safety. Unlike prior metrics, CRIMSON incorporates full clinical context, including patient age, indication, and guideline-based decision rules, and prevents normal or clinically insignificant findings from exerting disproportionate influence on the overall score. The framework categorizes errors into a comprehensive taxonomy covering false findings, missing findings, and eight attribute-level errors (e.g., location, severity, measurement, and diagnostic overinterpretation). Each finding is assigned a clinical significance level (urgent, actionable non-urgent, non-actionable, or expected/benign), based on a guideline developed in collaboration with attending cardiothoracic radiologists, enabling severity-aware weighting that prioritizes clinically consequential mistakes over benign discrepancies. CRIMSON is validated through strong alignment with clinically significant error counts annotated by six board-certified radiologists in ReXVal (Kendalls tau = 0.61-0.71; Pearsons r = 0.71-0.84), and through two additional benchmarks that we introduce. In RadJudge, a targeted suite of clinically challenging pass-fail scenarios, CRIMSON shows consistent agreement with expert judgment. In RadPref, a larger radiologist preference benchmark of over 100 pairwise cases with structured error categorization, severity modeling, and 1-5 overall quality ratings from three cardiothoracic radiologists, CRIMSON achieves the strongest alignment with radiologist preferences. We release the metric, the evaluation benchmarks, RadJudge and RadPref, and a fine-tuned MedGemma model to enable reproducible evaluation of report generation, all available at https://github.com/rajpurkarlab/CRIMSON.

Paperid: 25, https://arxiv.org/pdf/2603.06178.pdf GitHub

Abstract:
As powerful generative models, text-to-image diffusion models have recently been explored for discriminative tasks. A line of research focuses on adapting a pre-trained diffusion model to semantic segmentation without any further training, leading to what training-free diffusion segmentors. These methods typically rely on cross-attention maps from the model's attention layers, which are assumed to capture semantic relationships between image pixels and text tokens. Ideally, such approaches should benefit from more powerful diffusion models, i.e., stronger generative capability should lead to better segmentation. However, we observe that existing methods often fail to scale accordingly. To understand this issue, we identify two underlying gaps: (i) cross-attention is computed across multiple heads and layers, but there exists a discrepancy between these individual attention maps and a unified global representation. (ii) Even when a global map is available, it does not directly translate to accurate semantic correlation for segmentation, due to score imbalances among different text tokens. To bridge these gaps, we propose two techniques: auto aggregation and per-pixel rescaling, which together enable training-free segmentation to better leverage generative capability. We evaluate our approach on standard semantic segmentation benchmarks and further integrate it into a generative technique, demonstrating both improved performance broad applicability. Codes are at https://github.com/Darkbblue/goca.

Paperid: 26, https://arxiv.org/pdf/2603.06140.pdf GitHub

Abstract:
Modern video editing techniques have achieved high visual fidelity when inserting video objects. However, they focus on optimizing visual fidelity rather than physical causality, leading to edits that are physically inconsistent with their environment. In this work, we present Place-it-R$1$, an end-to-end framework for video object insertion that unlocks the environment-aware reasoning potential of Multimodal Large Language Models (MLLMs). Our framework leverages the Chain-of-Thought (CoT) reasoning of MLLMs to orchestrate video diffusion, following a Think-then-Place paradigm. To bridge cognitive reasoning and generative execution, we introduce three key innovations: First, MLLM performs physical scene understanding and interaction reasoning, generating environment-aware chain-of-thought tokens and inferring valid insertion regions to explicitly guide the diffusion toward physically plausible insertion. Then, we introduce MLLM-guided Spatial Direct Preference Optimization (DPO), where diffusion outputs are fed back to the MLLM for scoring, enabling visual naturalness. During inference, the MLLM iteratively triggers refinement cycles and elicits adaptive adjustments from the diffusion model, forming a closed-loop that progressively enhances editing quality. Furthermore, we provide two user-selectable modes: a plausibility-oriented flexible mode that permits environment modifications (\eg, generating support structures) to enhance physical plausibility, and a fidelity-oriented standard mode that preserves scene integrity for maximum fidelity, offering users explicit control over the plausibility-fidelity trade-off. Extensive experiments demonstrate Place-it-R1 achieves physically-coherent video object insertion compared with state-of-the-art solutions and commercial models.

Paperid: 27, https://arxiv.org/pdf/2603.06082.pdf GitHub

Abstract:
Recent advances in deep learning inspired neural network-based approaches to computational materials discovery (CMD). A plethora of problems in this field involve finding materials that optimize a target property. Nevertheless, the increasingly popular generative modeling methods are ineffective at boldly exploring attractive regions of the materials space due to their maximum likelihood training. In this work, we offer an alternative CMD technique based on offline model-based optimization (MBO) that fuses direct optimization of a target material property into generation. To that end, we introduce a domain-specific model, dubbed CliqueFlowmer, that incorporates recent advances of clique-based MBO into transformer and flow generation. We validate CliqueFlowmer's optimization abilities and show that materials it produces strongly outperform those provided by generative baselines. To enable employment of CliqueFlowmer in specialized materials optimization problems and support interdisciplinary research, we open-source our code at https://github.com/znowu/CliqueFlowmer.

Paperid: 28, https://arxiv.org/pdf/2603.06057.pdf GitHub

Abstract:
Diffusion models have recently advanced photorealistic human synthesis, although practical talking-head generation (THG) remains constrained by high inference latency, temporal instability such as flicker and identity drift, and imperfect audio-visual alignment under challenging speech conditions. This paper introduces TempoSyncDiff, a reference-conditioned latent diffusion framework that explores few-step inference for efficient audio-driven talking-head generation. The approach adopts a teacher-student distillation formulation in which a diffusion teacher trained with a standard noise prediction objective guides a lightweight student denoiser capable of operating with significantly fewer inference steps to improve generation stability. The framework incorporates identity anchoring and temporal regularization designed to mitigate identity drift and frame-to-frame flicker during synthesis, while viseme-based audio conditioning provides coarse lip motion control. Experiments on the LRS3 dataset report denoising-stage component-level metrics relative to VAE reconstructions and preliminary latency characterization, including CPU-only and edge computing measurements and feasibility estimates for edge deployment. The results suggest that distilled diffusion models can retain much of the reconstruction behaviour of a stronger teacher while enabling substantially lower latency inference. The study is positioned as an initial step toward practical diffusion-based talking-head generation under constrained computational settings. GitHub: https://mazumdarsoumya.github.io/TempoSyncDiff

Paperid: 29, https://arxiv.org/pdf/2603.06049.pdf GitHub

Abstract:
We identify a fundamental Narrow Policy limitation undermining the performance of autonomous VLA models, where driving Imitation Learning (IL) tends to collapse exploration and limit the potential of subsequent Reinforcement Learning (RL) stages, which often saturate prematurely due to insufficient feedback diversity. Thereby, we propose Curious-VLA, a framework that alleviates the exploit-explore dilemma through a two-stage design. During IL, we introduce a Feasible Trajectory Expansion (FTE) strategy to generate multiple physically valid trajectories and a step-wise normalized trajectory representation to adapt this diverse data. In the RL stage, we present Adaptive Diversity-Aware Sampling (ADAS) that prioritizes high-diversity samples and introduce Spanning Driving Reward (SDR) with a focal style weighting to amplify reward's value span for improving sensitivity to driving quality. On the Navsim benchmark, Curious-VLA achieves SoTA results (PDMS 90.3, EPDMS 85.4) and a Best-of-N PDMS of 94.8, demonstrating its effectiveness in unlocking the exploratory potential of VLA models. Code: https://github.com/Mashiroln/curious_vla.git.

Paperid: 30, https://arxiv.org/pdf/2603.06048.pdf GitHub

Abstract:
Hand-Object Interaction (HOI) remains a core challenge in digital human video synthesis, where models must generate physically plausible contact and preserve object identity across frames. Although recent HOI reenactment approaches have achieved progress, they are typically trained and evaluated in-domain and fail to generalize to complex, in-the-wild scenarios. In contrast, all-in-one video editing models exhibit broader robustness but still struggle with HOI-specific issues such as inconsistent object appearance. In this paper, we present GenHOI, a lightweight augmentation to pretrained video generation models that injects reference-object information in a temporally balanced and spatially selective manner. For temporal balancing, we propose Head-Sliding RoPE, which assigns head-specific temporal offsets to reference tokens, distributing their influence evenly across frames and mitigating the temporal decay of 3D RoPE to improve long-range object consistency. For spatial selectivity, we design a two-level spatial attention gate that concentrates object-conditioned attention on HOI regions and adaptively scales its strength, preserving background realism while enhancing interaction fidelity. Extensive qualitative and quantitative evaluations on unseen, in-the-wild scenes demonstrate that GenHOI significantly outperforms state-of-the-art HOI reenactment and all-in-one video editing methods. Project page: https://xuanhuang0.github.io/GenHOI/

Paperid: 31, https://arxiv.org/pdf/2603.06038.pdf GitHub

Abstract:
Recent text-to-image models can generate high-quality images from natural-language prompts, yet controlling typography remains challenging: requested typographic appearance is often ignored or only weakly followed. We address this limitation with a data-centric approach that trains image generation models using targeted supervision derived from a structured annotation pipeline specialized for typography. Our pipeline constructs a large-scale typography-focused dataset, FontUse, consisting of about 70K images annotated with user-friendly prompts, text-region locations, and OCR-recognized strings. The annotations are automatically produced using segmentation models and multimodal large language models (MLLMs). The prompts explicitly combine font styles (e.g., serif, script, elegant) and use cases (e.g., wedding invitations, coffee-shop menus), enabling intuitive specification even for novice users. Fine-tuning existing generators with these annotations allows them to consistently interpret style and use-case conditions as textual prompts without architectural modification. For evaluation, we introduce a Long-CLIP-based metric that measures alignment between generated typography and requested attributes. Experiments across diverse prompts and layouts show that models trained with our pipeline produce text renderings more consistent with prompts than competitive baselines. The source code for our annotation pipeline is available at https://github.com/xiaxinz/FontUSE.

Paperid: 32, https://arxiv.org/pdf/2603.06007.pdf GitHub

Abstract:
Large language model-based (LLM-based) multi-agent systems (MAS) are increasingly used to extend agentic problem solving via role specialization and collaboration. MAS workflows can be naturally modeled as directed computation graphs, where nodes execute agents/sub-workflows and edges encode dependencies and message passing. However, implementing complex graph workflows in current frameworks still requires substantial manual effort, offers limited reuse, and makes it difficult to integrate heterogeneous external context sources. To overcome these limitations, we present MASFactory, a graph-centric framework for orchestrating LLM-based MAS. It introduces Vibe Graphing, a human-in-the-loop approach that compiles natural-language intent into an editable workflow specification and then into an executable graph. In addition, the framework provides reusable components and pluggable context integration, as well as a visualizer for topology preview, runtime tracing, and human-in-the-loop interaction. We evaluate MASFactory on seven public benchmarks, validating both reproduction consistency for representative MAS methods and the effectiveness of Vibe Graphing. Our code (https://github.com/BUPT-GAMMA/MASFactory) and video (https://youtu.be/ANynzVfY32k) are publicly available.

Paperid: 33, https://arxiv.org/pdf/2603.05996.pdf GitHub

Abstract:
Generative language models have shown significant potential in single-turn Text-to-SQL. However, their performance does not extend equivalently to multi-turn Text-to-SQL. This is primarily due to generative language models' inadequacy in handling the complexities of context information and dynamic schema linking in multi-turn interactions. In this paper, we propose a framework named Track-SQL, which enhances generative language models with dual-extractive modules designed to track schema and contextual changes in multi-turn Text-to-SQL. Specifically, Track-SQL incorporates a \emph{Semantic-enhanced Schema Extractor} and a \emph{Schema-aware Context Extractor}. Experimental results demonstrate that Track-SQL achieves state-of-the-art performance on the SparC and CoSQL datasets. Furthermore, detailed ablation studies reveal that Track-SQL significantly improves execution accuracy in multi-turn interactions by 7.1\% and 9.55\% on these datasets, respectively. Our implementation will be open-sourced at https://github.com/DMIRLAB-Group/Track-SQL.

Paperid: 34, https://arxiv.org/pdf/2603.05972.pdf GitHub

Abstract:
The explosion of big social data has created a scalability trap for traditional qualitative research, as manual coding remains labor-intensive and conventional topic models often suffer from semantic thinning and a lack of domain awareness. This paper introduces Textual Hybrid Embedding based Topic Analysis (THETA), a novel computational paradigm and open-source tool designed to bridge the gap between massive data scale and rich theoretical depth. THETA moves beyond frequency-based statistics by implementing Domain-Adaptive Fine-tuning (DAFT) via LoRA on foundation embedding models, which effectively optimizes semantic vector structures within specific social contexts to capture latent meanings. To ensure epistemological rigor, we encapsulate this process into an AI Scientist Agent framework, comprising Data Steward, Modeling Analyst, and Domain Expert agents, to simulate the human-in-the-loop expert judgment and constant comparison processes central to grounded theory. Departing from purely computational models, this framework enables agents to iteratively evaluate algorithmic clusters, perform cross-topic semantic alignment, and refine raw outputs into logically consistent theoretical categories. To validate the effectiveness of THETA, we conducted experiments across six domains, including financial regulation and public health. Our results demonstrate that THETA significantly outperforms traditional models, such as LDA, ETM, and CTM, in capturing domain-specific interpretive constructs while maintaining superior coherence. By providing an interactive analysis platform, THETA democratizes advanced natural language processing for social scientists and ensures the trustworthiness and reproducibility of research findings. Code is available at https://github.com/CodeSoul-co/THETA.

Paperid: 35, https://arxiv.org/pdf/2603.05969.pdf GitHub

Abstract:
Change captioning generates descriptions that explicitly describe the differences between two visually similar images. Existing methods operate on static image pairs, thus ignoring the rich temporal dynamics of the change procedure, which is the key to understand not only what has changed but also how it occurs. We introduce ProCap, a novel framework that reformulates change modeling from static image comparison to dynamic procedure modeling. ProCap features a two-stage design: The first stage trains a procedure encoder to learn the change procedure from a sparse set of keyframes. These keyframes are obtained by automatically generating intermediate frames to make the implicit procedural dynamics explicit and then sampling them to mitigate redundancy. Then the encoder learns to capture the latent dynamics of these keyframes via a caption-conditioned, masked reconstruction task. The second stage integrates this trained encoder within an encoder-decoder model for captioning. Instead of relying on explicit frames from the previous stage -- a process incurring computational overhead and sensitivity to visual noise -- we introduce learnable procedure queries to prompt the encoder for inferring the latent procedure representation, which the decoder then translates into text. The entire model is then trained end-to-end with a captioning loss, ensuring the encoder's output is both temporally coherent and captioning-aligned. Experiments on three datasets demonstrate the effectiveness of ProCap. Code and pre-trained models are available at https://github.com/BlueberryOreo/ProCap

Paperid: 36, https://arxiv.org/pdf/2603.05952.pdf GitHub

Abstract:
Few-shot segmentation (FSS) has gained significant attention for its ability to generalize to novel classes with limited supervision, yet remains challenged by structural misalignment and cross-view inconsistency under large appearance or viewpoint variations. This paper tackles these challenges by introducing VINE (View-Informed NEtwork), a unified framework that jointly models structural consistency and foreground discrimination to refine class-specific prototypes. Specifically, VINE introduces a spatial-view graph on backbone features, where the spatial graph captures local geometric topology and the view graph connects features from different perspectives to propagate view-invariant structural semantics. To further alleviate foreground ambiguity, we derive a discriminative prior from the support-query feature discrepancy to capture category-specific contrast, which reweights SAM features by emphasizing salient regions and recalibrates backbone activations for improved structural focus. The foreground-enhanced SAM features and structurally enriched ResNet features are progressively integrated through masked cross-attention, yielding class-consistent prototypes used as adaptive prompts for the SAM decoder to generate accurate masks. Extensive experiments on multiple FSS benchmarks validate the effectiveness and robustness of VINE, particularly under challenging scenarios with viewpoint shifts and complex structures. The code is available at https://github.com/HongliLiu1/VINE-main.

Paperid: 37, https://arxiv.org/pdf/2603.05942.pdf GitHub

Abstract:
Skew estimation is one of the vital tasks in document processing systems, especially for scanned document images, because its performance impacts subsequent steps directly. Over the years, an enormous number of researches focus on this challenging problem in the rise of digitization age. In this research, we first propose a novel skew estimation method that extracts the dominant skew angle of the given document image by applying an Adaptive Radial Projection on the 2D Discrete Fourier Magnitude spectrum. Second, we introduce a high quality skew estimation dataset DISE-2021 to assess the performance of different estimators. Finally, we provide comprehensive analyses that focus on multiple improvement aspects of Fourier-based methods. Our results show that the proposed method is robust, reliable, and outperforms all compared methods. The source code is available at https://github.com/phamquiluan/jdeskew.

Paperid: 38, https://arxiv.org/pdf/2603.05937.pdf GitHub

Abstract:
Automatic facial expression recognition (FER) has gained much attention due to its applications in human-computer interaction. Among the approaches to improve FER tasks, this paper focuses on deep architecture with the attention mechanism. We propose a novel Masking idea to boost the performance of CNN in facial expression task. It uses a segmentation network to refine feature maps, enabling the network to focus on relevant information to make correct decisions. In experiments, we combine the ubiquitous Deep Residual Network and Unet-like architecture to produce a Residual Masking Network. The proposed method holds state-of-the-art (SOTA) accuracy on the well-known FER2013 and private VEMO datasets. The source code is available at https://github.com/phamquiluan/ResidualMaskingNetwork.

Paperid: 39, https://arxiv.org/pdf/2603.05935.pdf GitHub

Abstract:
High-speed aerial grasping presents significant challenges due to the high demands on precise, responsive flight control and coordinated gripper manipulation. In this work, we propose Swooper, a deep reinforcement learning (DRL) based approach that achieves both precise flight control and active gripper control using a single lightweight neural network policy. Training such a policy directly via DRL is nontrivial due to the complexity of coordinating flight and grasping. To address this, we adopt a two-stage learning strategy: we first pre-train a flight control policy, and then fine-tune it to acquire grasping skills. With the carefully designed reward functions and training framework, the entire training process completes in under 60 minutes on a standard desktop with an Nvidia RTX 3060 GPU. To validate the trained policy in the real world, we develop a lightweight quadrotor grasping platform equipped with a simple off-the-shelf gripper, and deploy the policy in a zero-shot manner on the onboard Raspberry Pi 4B computer, where each inference takes only about 1.0 ms. In 25 real-world trials, our policy achieves an 84% grasp success rate and grasping speeds of up to 1.5 m/s without any fine-tuning. This matches the robustness and agility of state-of-the-art classical systems with sophisticated grippers, highlighting the capability of DRL for learning a robust control policy that seamlessly integrates high-speed flight and grasping. The supplementary video is available for more results. Video: https://zikenhuang.github.io/Swooper/.

Paperid: 40, https://arxiv.org/pdf/2603.05929.pdf GitHub

Abstract:
Vision Transformers (ViTs) have recently achieved state-of-the-art performance in 2D human pose estimation due to their strong global modeling capability. However, existing ViT-based pose estimators are designed for static images and process each frame independently, thereby ignoring the temporal coherence that exists in video sequences. This limitation often results in unstable predictions, especially in challenging scenes involving motion blur, occlusion, or defocus. In this paper, we propose TAR-ViTPose, a novel Temporal Aggregate-and-Restore Vision Transformer tailored for video-based 2D human pose estimation. TAR-ViTPose enhances static ViT representations by aggregating temporal cues across frames in a plug-and-play manner, leading to more robust and accurate pose estimation. To effectively aggregate joint-specific features that are temporally aligned across frames, we introduce a joint-centric temporal aggregation (JTA) that assigns each joint a learnable query token to selectively attend to its corresponding regions from neighboring frames. Furthermore, we develop a global restoring attention (GRA) to restore the aggregated temporal features back into the token sequence of the current frame, enriching its pose representation while fully preserving global context for precise keypoint localization. Extensive experiments demonstrate that TAR-ViTPose substantially improves upon the single-frame baseline ViTPose, achieving a +2.3 mAP gain on the PoseTrack2017 benchmark. Moreover, our approach outperforms existing state-of-the-art video-based methods, while also achieving a noticeably higher runtime frame rate in real-world applications. Project page: https://github.com/zgspose/TARViTPose.

Paperid: 41, https://arxiv.org/pdf/2603.05925.pdf GitHub

Abstract:
In this paper, we propose a Rectified Flow Auto Coder (RAC) inspired by Rectified Flow to replace the traditional VAE: 1. It achieves multi-step decoding by applying the decoder to flow timesteps. Its decoding path is straight and correctable, enabling step-by-step refinement. 2. The model inherently supports bidirectional inference, where the decoder serves as the encoder through time reversal (hence Coder rather than encoder or decoder), reducing parameter count by nearly 41%. 3. This generative decoding method improves generation quality since the model can correct latent variables along the path, partially addressing the reconstruction--generation gap. Experiments show that RAC surpasses SOTA VAEs in both reconstruction and generation with approximately 70% lower computational cost.

Paperid: 42, https://arxiv.org/pdf/2603.05924.pdf GitHub GitHub

Abstract:
Modern neural network optimization relies heavily on architectural priorssuch as Batch Normalization and Residual connectionsto stabilize training dynamics. Without these, or in low-data regimes with aggressive augmentation, low-bias architectures like Vision Transformers (ViTs) often suffer from optimization collapse. This work adopts Sketched Isotropic Gaussian Regularization (SIGReg), recently introduced in the LeJEPA self-supervised framework, and repurposes it as a general optimization stabilizer for supervised learning. While the original formulation targets the full characteristic function, a computationally efficient variant is derived, Weak-SIGReg, which targets the covariance matrix via random sketching. Inspired by interacting particle systems, representation collapse is viewed as stochastic drift; SIGReg constrains the representation density towards an isotropic Gaussian, mitigating this drift. Empirically, SIGReg recovers the training of a ViT on CIFAR-100 from a collapsed 20.73\% to 72.02\% accuracy without architectural hacks and significantly improves the convergence of deep vanilla MLPs trained with pure SGD. Code is available at \href{https://github.com/kreasof-ai/sigreg}{github.com/kreasof-ai/sigreg}.

Paperid: 43, https://arxiv.org/pdf/2603.05923.pdf GitHub

Abstract:
Truly proactive AI systems must anticipate what we will do next. This foresight demands far richer information than the sparse signals we type into our prompts -- it demands reasoning over the entire context of what we see and do. We formalize this as next action prediction (NAP): given a sequence of a user's multimodal interactions with a computer (screenshots, clicks, sensor data), predict that user's next action. Progress on this task requires both new data and modeling approaches. To scale data, we annotate longitudinal, naturalistic computer use with vision-language models. We release an open-source pipeline for performing this labeling on private infrastructure, and label over 360K actions across one month of continuous phone usage from 20 users, amounting to 1,800 hours of screen time. We then introduce LongNAP, a user model that combines parametric and in-context learning to reason over long interaction histories. LongNAP is trained via policy gradient methods to generate user-specific reasoning traces given some context; retrieve relevant traces from a library of past traces; and then apply retrieved traces in-context to predict future actions. Using an LLM-as-judge evaluation metric (0-1 similarity to ground truth), LongNAP significantly outperforms supervised finetuning and prompted baselines on held-out data (by 79% and 39% respectively). Additionally, LongNAP generalizes to held out users when trained across individuals. The space of next actions a user might take at any moment is unbounded, spanning thousands of possible outcomes. Despite this, 17.1% of LongNAP's predicted trajectories are well-aligned with what a user does next (LLM-judge score $\geq$ 0.5). This rises to 26% when we filter to highly confident predictions. In sum, we argue that learning from the full context of user behavior to anticipate user needs is now a viable task with substantial opportunity.

Paperid: 44, https://arxiv.org/pdf/2603.05921.pdf GitHub

Abstract:
This paper investigates the challenging task of detecting backdoored text-to-image models under black-box settings and introduces a novel detection framework BlackMirror. Existing approaches typically rely on analyzing image-level similarity, under the assumption that backdoor-triggered generations exhibit strong consistency across samples. However, they struggle to generalize to recently emerging backdoor attacks, where backdoored generations can appear visually diverse. BlackMirror is motivated by an observation: across backdoor attacks, {only partial semantic patterns within the generated image are steadily manipulated, while the rest of the content remains diverse or benign. Accordingly, BlackMirror consists of two components: MirrorMatch, which aligns visual patterns with the corresponding instructions to detect semantic deviations; and MirrorVerify, which evaluates the stability of these deviations across varied prompts to distinguish true backdoor behavior from benign responses. BlackMirror is a general, training-free framework that can be deployed as a plug-and-play module in Model-as-a-Service (MaaS) applications. Comprehensive experiments demonstrate that BlackMirror achieves accurate detection across a wide range of attacks. Code is available at https://github.com/Ferry-Li/BlackMirror.

Paperid: 45, https://arxiv.org/pdf/2603.05911.pdf GitHub

Abstract:
Medical image segmentation is undergoing a paradigm shift from conventional visual pattern matching to cognitive reasoning analysis. Although Multimodal Large Language Models (MLLMs) have shown promise in integrating linguistic and visual knowledge, significant gaps remain: existing general MLLMs possess broad common sense but lack the specialized visual reasoning required for complex lesions, whereas traditional segmentation models excel at pixel-level segmentation but lack logical interpretability. In this paper, we introduce ComLesion-14K, the first diverse Chain-of-Thought (CoT) benchmark for reasoning-driven complex lesion segmentation. To accomplish this task, we propose CORE-Seg, an end-to-end framework integrating reasoning with segmentation through a Semantic-Guided Prompt Adapter. We design a progressive training strategy from SFT to GRPO, equipped with an adaptive dual-granularity reward mechanism to mitigate reward sparsity. Our Method achieves state-of-the-art results with a mean Dice of 37.06\% (14.89\% higher than the second-best baseline), while reducing the failure rate to 18.42\%. Project Page: https://xyxl024.github.io/CORE-Seg.github.io/

Paperid: 46, https://arxiv.org/pdf/2603.05908.pdf GitHub GitHub

Abstract:
Current compositional image-to-3D scene generation approaches construct 3D scenes by time-consuming iterative layout optimization or inflexible joint object-layout generation. Moreover, most methods rely on limited field-of-view perspective images, hindering the creation of complete 360-degree environments. To address these limitations, we design Pano3DComposer, an efficient feed-forward framework for panoramic images. To decouple object generation from layout estimation, we propose a plug-and-play Object-World Transformation Predictor. This module converts the 3D objects generated by off-the-shelf image-to-3D models from local to world coordinates. To achieve this, we adapt the VGGT architecture to Alignment-VGGT by using target object crop, multi-view object renderings and camera parameters to predict the transformation. The predictor is trained using pseudo-geometric supervision to address the shape discrepancy between generated and ground-truth objects. For input images from unseen domains, we further introduce a Coarse-to-Fine (C2F) alignment mechanism for Pano3DComposer that iteratively refines geometric consistency with feedback of scene rendering. Our method achieves superior geometric accuracy for image/text-to-3D tasks on synthetic and real-world datasets. It can generate a high-fidelity 3D scene in approximately 20 seconds on an RTX 4090 GPU. Project page: https://qiuzidian.github.io/pano3dcomposer-page/.

Paperid: 47, https://arxiv.org/pdf/2603.05905.pdf GitHub

Abstract:
Small object detection in unmanned aerial vehicle (UAV) imagery is challenging, mainly due to scale variation, structural detail degradation, and limited computational resources. In high-altitude scenarios, fine-grained features are further weakened during hierarchical downsampling and cross-scale fusion, resulting in unstable localization and reduced robustness. To address this issue, we propose CollabOD, a lightweight collaborative detection framework that explicitly preserves structural details and aligns heterogeneous feature streams before multi-scale fusion. The framework integrates Structural Detail Preservation, Cross-Path Feature Alignment, and Localization-Aware Lightweight Design strategies. From the perspectives of image processing, channel structure, and lightweight design, it optimizes the architecture of conventional UAV perception models. The proposed design enhances representation stability while maintaining efficient inference. A unified detail-aware detection head further improves regression robustness without introducing additional deployment overhead. The code is available at: https://github.com/Bai-Xuecheng/CollabOD.

Paperid: 48, https://arxiv.org/pdf/2603.05900.pdf GitHub

Abstract:
Large language models (LLMs) benefit substantially from supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR) in reasoning tasks. However, these recipes perform poorly in instruction-based molecular optimization, where each data point typically provides only a single optimized reference molecule and no step-by-step optimization trajectory. We reveal that answer-only SFT on the reference molecules collapses reasoning, and RLVR provides sparse feedback under similarity constraints due to the model's lack of effective exploration, which slows learning and limits optimization. To encourage the exploration of new molecules while balancing the exploitation of the reference molecules, we introduce Reference-guided Policy Optimization (RePO), an optimization approach that learns from reference molecules without requiring trajectory data. At each update, RePO samples candidate molecules with their intermediate reasoning trajectories from the model and trains the model using verifiable rewards that measure property satisfaction under similarity constraints in an RL manner. Meanwhile, it applies reference guidance by keeping the policy's intermediate reasoning trajectory as context and training only the answer in a supervised manner. Together, the RL term promotes exploration, while the guidance term mitigates reward sparsity and stabilizes training by grounding outputs to references when many valid molecular edits exist. Across molecular optimization benchmarks, RePO consistently outperforms SFT and RLVR baselines (e.g., GRPO), achieving improvements on the optimization metric (Success Rate $\times$ Similarity), improving balance across competing objectives, and generalizing better to unseen instruction styles. Our code is publicly available at https://github.com/tmlr-group/RePO.

Paperid: 49, https://arxiv.org/pdf/2603.05890.pdf GitHub

Abstract:
What happens when a storyteller forgets its own story? Large Language Models (LLMs) can now generate narratives spanning tens of thousands of words, but they often fail to maintain consistency throughout. When generating long-form narratives, these models can contradict their own established facts, character traits, and world rules. Existing story generation benchmarks focus mainly on plot quality and fluency, leaving consistency errors largely unexplored. To address this gap, we present ConStory-Bench, a benchmark designed to evaluate narrative consistency in long-form story generation. It contains 2,000 prompts across four task scenarios and defines a taxonomy of five error categories with 19 fine-grained subtypes. We also develop ConStory-Checker, an automated pipeline that detects contradictions and grounds each judgment in explicit textual evidence. Evaluating a range of LLMs through five research questions, we find that consistency errors show clear tendencies: they are most common in factual and temporal dimensions, tend to appear around the middle of narratives, occur in text segments with higher token-level entropy, and certain error types tend to co-occur. These findings can inform future efforts to improve consistency in long-form narrative generation. Our project page is available at https://picrew.github.io/constory-bench.github.io/.

Paperid: 50, https://arxiv.org/pdf/2603.05888.pdf GitHub

Abstract:
We introduce PixARMesh, a method to autoregressively reconstruct complete 3D indoor scene meshes directly from a single RGB image. Unlike prior methods that rely on implicit signed distance fields and post-hoc layout optimization, PixARMesh jointly predicts object layout and geometry within a unified model, producing coherent and artist-ready meshes in a single forward pass. Building on recent advances in mesh generative models, we augment a point-cloud encoder with pixel-aligned image features and global scene context via cross-attention, enabling accurate spatial reasoning from a single image. Scenes are generated autoregressively from a unified token stream containing context, pose, and mesh, yielding compact meshes with high-fidelity geometry. Experiments on synthetic and real-world datasets show that PixARMesh achieves state-of-the-art reconstruction quality while producing lightweight, high-quality meshes ready for downstream applications.

Paperid: 51, https://arxiv.org/pdf/2603.05887.pdf GitHub

Abstract:
Neural audio codecs optimized for mel-spectrogram reconstruction often fail to preserve intelligibility. While semantic encoder distillation improves encoded representations, it does not guarantee content preservation in reconstructed speech. In this work, we demonstrate that self-supervised representation reconstruction (SSRR) loss fundamentally improves codec training and performance. First, SSRR significantly accelerates convergence, enabling competitive results using only a single GPU. Second, it enhances intelligibility by reconstructing distilled self-supervised representations from codec outputs. Third, SSRR enables high intelligibility without additional lookahead in streaming Transformer-based codecs, allowing a zero-lookahead architecture for real-time deployment. As a result, our JHCodec achieves state-of-the-art performance while maintaining minimal latency and reduced training cost. We open-source the full implementation, training pipeline, and demo on Github https://github.com/jhcodec843/jhcodec.

Paperid: 52, https://arxiv.org/pdf/2603.05878.pdf GitHub

Abstract:
Pruning is widely recognized as an effective method for reducing the parameters of large language models (LLMs), potentially leading to more efficient deployment and inference. One classic and prominent path of LLM one-shot pruning is to leverage second-order gradients (i.e., Hessian), represented by the pioneering work SparseGPT. However, the predefined left-to-right pruning order in SparseGPT leads to suboptimal performance when the weights exhibit columnar patterns. This paper studies the effect of pruning order under the SparseGPT framework. The analyses lead us to propose ROSE, a reordered SparseGPT method that prioritizes weights with larger potential pruning errors to be pruned earlier. ROSE first performs pre-pruning to identify candidate weights for removal, and estimates both column and block pruning loss. Subsequently, two-level reordering is performed: columns within each block are reordered in descending order of column loss, while blocks are reordered based on block loss. We introduce the relative range of block loss as a metric to identify columnar layers, enabling adaptive reordering across the entire model. Substantial empirical results on prevalent LLMs (LLaMA2-7B/13B/70B, LLaMA3-8B, Mistral-7B) demonstrate that ROSE surpasses the original SparseGPT and other counterpart pruning methods. Our code is available at https://github.com/mingluo-su/ROSE.

Paperid: 53, https://arxiv.org/pdf/2603.05868.pdf GitHub

Abstract:
Despite remarkable progress in Vision-Language-Action models (VLAs) for robot manipulation, these large pre-trained models require fine-tuning to be deployed in specific environments. These fine-tuned models are highly sensitive to camera viewpoint changes that frequently occur in unstructured environments. In this paper, we propose a zero-shot camera adaptation framework without additional demonstration data, policy fine-tuning, or architectural modification. Our key idea is to virtually adjust test-time camera observations to match the training camera configuration in real-time. For that, we use a recent feed-forward novel view synthesis model which outputs high-quality target view images, handling both extrinsic and intrinsic parameters. This plug-and-play approach preserves the pre-trained capabilities of VLAs and applies to any RGB-based policy. Through extensive experiments on the LIBERO benchmark, our method consistently outperforms baselines that use data augmentation for policy fine-tuning or additional 3D-aware features for visual input. We further validate that our approach constantly enhances viewpoint robustness in real-world robotic manipulation scenarios, including settings with varying camera extrinsics, intrinsics, and freely moving handheld cameras.

Paperid: 54, https://arxiv.org/pdf/2603.05867.pdf GitHub

Abstract:
Accurate tumor analysis is central to clinical radiology and precision oncology, where early detection, reliable lesion characterization, and pathology-level risk assessment guide diagnosis and treatment planning. Chain-of-Thought (CoT) reasoning is particularly important in this setting because it enables step-by-step interpretation from imaging findings to clinical impressions and pathology conclusions, improving traceability and reducing diagnostic errors. Here, we target the clinical tumor analysis task and build a large-scale benchmark that operationalizes a multimodal reasoning pipeline, spanning findings, impressions, and pathology predictions. We curate TumorCoT, a large-scale dataset of 1.5M CoT-labeled VQA instructions paired with 3D CT scans, with step-aligned rationales and cross-modal alignments along the trajectory from findings to impression to pathology, enabling evaluation of both answer accuracy and reasoning consistency. We further propose TumorChain, a multimodal interleaved reasoning framework that tightly couples 3D imaging encoders, clinical text understanding, and organ-level vision-language alignment. Through cross-modal alignment and iterative interleaved causal reasoning, TumorChain grounds visual evidence, aggregates conclusions, and issues pathology predictions after multiple rounds of self-refinement, improving traceability and reducing hallucination risk. Experiments show consistent improvements over strong baselines in lesion detection, impression generation, and pathology classification, and demonstrate strong generalization on the DeepTumorVQA benchmark. These results highlight the potential of multimodal reasoning for reliable and interpretable tumor analysis in clinical practice. Detailed information about our project can be found on our project homepage at https://github.com/ZJU4HealthCare/TumorChain.

Paperid: 55, https://arxiv.org/pdf/2603.05863.pdf GitHub

Abstract:
While Large Language Models (LLMs) have revolutionized code generation, standard "System 1" approaches, generating solutions in a single forward pass, often hit a performance ceiling when faced with complex algorithmic tasks. Existing iterative refinement strategies attempt to bridge this gap at inference time, yet they predominantly rely on external oracles, execution feedback, or computationally expensive prompt-response cycles. In this work, we propose ReflexiCoder, a novel reinforcement learning (RL) framework that internalizes the structured reasoning trajectory, encompassing initial generation, bug and optimization aware reflection, and self-correction, directly into the model's weights. Unlike prior methods, ReflexiCoder shifts the paradigm from external-dependent refinement to an intrinsic, fully autonomous self-reflection and self-correction capabilities at inference time. We utilize an RL-zero training paradigm with granular reward functions to optimize the entire reflection-correction trajectory, teaching the model how to debug without reliance on ground-truth feedback or execution engines at inference time. Extensive experiments across seven benchmarks demonstrate that our ReflexiCoder-8B establishes a new state-of-the-art (SOTA) among leading open-source models in the 1.5B-14B range, achieving 94.51% (87.20%) on HumanEval (Plus), 81.80% (78.57%) on MBPP (Plus), 35.00% on BigCodeBench, 52.21% on LiveCodeBench, and 37.34% on CodeForces in a single-attempt setting, rivaling or surpassing proprietary models like GPT-5.1. Notably, our framework is significantly more token-efficient than base models, reducing inference-time compute overhead by approximately 40% through disciplined, high-speed reasoning and reflection patterns. Source code is available at https://github.com/juyongjiang/ReflexiCoder.

Paperid: 56, https://arxiv.org/pdf/2603.05830.pdf GitHub

Abstract:
Legged manipulators offer high mobility and versatile manipulation. However, robust interaction with heterogeneous articulated objects, such as doors, drawers, and cabinets, remains challenging because of the diverse articulation types of the objects and the complex dynamics of the legged robot. Existing reinforcement learning (RL)-based approaches often rely on high-dimensional sensory inputs, leading to sample inefficiency. In this paper, we propose a robust and sample-efficient framework for opening heterogeneous articulated objects with a legged manipulator. In particular, we propose Sampling-based Abstracted Feature Extraction (SAFE), which encodes handle and panel geometry into a compact low-dimensional representation, improving cross-domain generalization. Additionally, Articulation Information Estimator (ArtIEst) is introduced to adaptively mix proprioception with exteroception to estimate opening direction and range of motion for each object. The proposed framework was deployed to manipulate various heterogeneous articulated objects in simulation and real-world robot systems. Videos can be found on the project website: https://openheart-icra.github.io/OpenHEART/

Paperid: 57, https://arxiv.org/pdf/2603.05822.pdf GitHub

Abstract:
Adapting foundation models to new clinical sites remains challenging in practice. Domain shift and scarce annotations must be handled by experts, yet many clinical groups do not have ready access to skilled AI engineers to tune adapter designs and training recipes. As a result, adaptation cycles can stretch from weeks to months, particularly in few-shot settings. Existing PEFT methods either require manual adapter configuration or automated searches that are computationally infeasible in few-shot 3D settings. We propose SEA-PEFT (SElf-Auditing Parameter-Efficient Fine-Tuning) to automate this process. SEA-PEFT treats adapter configuration as an online allocation problem solved during fine-tuning rather than through manual, fixed-topology choices. SEA-PEFT uses a search-audit-allocate loop that trains active adapters, estimates each adapter's Dice utility by momentarily toggling it off, and then reselects the active set under a parameter budget using a greedy knapsack allocator. Exponential Moving Average and Interquartile Range smoothing, together with a Finite-State Ranking controller, stabilize the loop and improve reliability in high-noise few-shot regimes. On TotalSegmentator and FLARE'22, SEA-PEFT improves mean Dice by 2.4--2.8 points over the strongest fixed-topology PEFT baselines across 1/5/10-shot settings while training <1% of parameters. For reproducibility purposes, we made our code publicly available at https://github.com/tsly123/SEA_PEFT

Paperid: 58, https://arxiv.org/pdf/2603.05804.pdf GitHub

Abstract:
High-quality teleoperated demonstrations are a primary bottleneck for imitation learning (IL) in dexterous manipulation. However, haptic feedback provides operators with real-time contact information, enabling real-time finger posture adjustments, and thereby improving demonstration quality. Existing dexterous teleoperation platforms typically omit haptic feedback and remain bulky and expensive. We introduce CDF-Glove, a lightweight and low cost cable-driven force-feedback glove. The real-time state is available for 20 finger degrees of freedom (DoF), of which 16 are directly sensed and 4 are passively coupled (inferred from kinematic constraints). We develop a kinematic model and control stack for the glove, and validate them across multiple robotic hands with diverse kinematics and DoF. The CDF-Glove achieves distal joint repeatability of 0.4 degrees, and delivers about 200 ms force feedback latency, yielding a 4x improvement in task success rate relative to no-feedback teleoperation. We collect two bimanual teleoperation datasets, on which we train and evaluate Diffusion Policy baselines. Compared to kinesthetic teaching, the policies trained in our teleoperated demonstrations increase the average success rate by 55% and reduce the mean completion time by approximately 15.2 seconds (a 47.2% relative reduction). In particular, the CDF-Glove costs approximately US$230. The code and designs are released as open source at https://cdfglove.github.io/.

Paperid: 59, https://arxiv.org/pdf/2603.05786.pdf GitHub

Abstract:
As AI agents become widely deployed as online services, users often rely on an agent developer's claim about how safety is enforced, which introduces a threat where safety measures are falsely advertised. To address the threat, we propose proof-of-guardrail, a system that enables developers to provide cryptographic proof that a response is generated after a specific open-source guardrail. To generate proof, the developer runs the agent and guardrail inside a Trusted Execution Environment (TEE), which produces a TEE-signed attestation of guardrail code execution verifiable by any user offline. We implement proof-of-guardrail for OpenClaw agents and evaluate latency overhead and deployment cost. Proof-of-guardrail ensures integrity of guardrail execution while keeping the developer's agent private, but we also highlight a risk of deception about safety, for example, when malicious developers actively jailbreak the guardrail. Code and demo video: https://github.com/SaharaLabsAI/Verifiable-ClawGuard

Paperid: 60, https://arxiv.org/pdf/2603.05767.pdf GitHub

Abstract:
We address multi-robot motion planning under Signal Temporal Logic (STL) specifications with kinodynamic constraints. Exact approaches face scalability bottlenecks and limited adaptability, while conventional sampling-based methods require excessive samples to construct optimal trajectories. We propose a two-stage framework integrating sampling-based online learning with formal STL reasoning. At the single-robot level, our constrained Bayesian Optimization-based Tree search (cBOT) planner uses a Gaussian process as a surrogate model to learn local cost maps and feasibility constraints, generating shorter collision-free trajectories with fewer samples. At the multi-robot level, our STL-enhanced Kinodynamic Conflict-Based Search (STL-KCBS) algorithm incorporates STL monitoring into conflict detection and resolution, ensuring specification satisfaction while maintaining scalability and probabilistic completeness. Benchmarking demonstrates improved trajectory efficiency and safety over existing methods. Real-world experiments with autonomous surface vehicles validate robustness and practical applicability in uncertain environments. The STLcBOT Planner will be released as an open-source package, and videos of real-world and simulated experiments are available at https://stlbot.github.io/.

Paperid: 61, https://arxiv.org/pdf/2603.05764.pdf GitHub

Abstract:
Autonomous coding agents can produce strong tabular baselines quickly on Kaggle-style tasks. Practical value depends on end-to-end correctness and reliability under time limits. This paper introduces TML-Bench, a tabular benchmark for data science agents on Kaggle-style tasks. This paper evaluates 10 OSS LLMs on four Kaggle competitions and three time budgets (240s, 600s, and 1200s). Each model is run five times per task and budget. A run is successful if it produces a valid submission and a private-holdout score on hidden labels that are not accessible to the agent. This paper reports median performance, success rates, and run-to-run variability. MiniMax-M2.1 model achieves the best aggregate performance score on all four competitions under the paper's primary aggregation. Average performance improves with larger time budgets. Scaling is noisy for some individual models at the current run count. Code and materials are available at https://github.com/MykolaPinchuk/TML-bench/tree/master.

Paperid: 62, https://arxiv.org/pdf/2603.05729.pdf GitHub

Abstract:
The original ImageNet benchmark enforces a single-label assumption, despite many images depicting multiple objects. This leads to label noise and limits the richness of the learning signal. Multi-label annotations more accurately reflect real-world visual scenes, where multiple objects co-occur and contribute to semantic understanding, enabling models to learn richer and more robust representations. While prior efforts (e.g., ReaL, ImageNetv2) have improved the validation set, there has not yet been a scalable, high-quality multi-label annotation for the training set. To this end, we present an automated pipeline to convert the ImageNet training set into a multi-label dataset, without human annotations. Using self-supervised Vision Transformers, we perform unsupervised object discovery, select regions aligned with original labels to train a lightweight classifier, and apply it to all regions to generate coherent multi-label annotations across the dataset. Our labels show strong alignment with human judgment in qualitative evaluations and consistently improve performance across quantitative benchmarks. Compared to traditional single-label scheme, models trained with our multi-label supervision achieve consistently better in-domain accuracy across architectures (up to +2.0 top-1 accuracy on ReaL and +1.5 on ImageNet-V2) and exhibit stronger transferability to downstream tasks (up to +4.2 and +2.3 mAP on COCO and VOC, respectively). These results underscore the importance of accurate multi-label annotations for enhancing both classification performance and representation learning. Project code and the generated multi-label annotations are available at https://github.com/jchen175/MultiLabel-ImageNet.

Paperid: 63, https://arxiv.org/pdf/2603.05711.pdf GitHub

Abstract:
Accurate, dense depth estimation is crucial for robotic perception, but commodity sensors often yield sparse or incomplete measurements due to hardware limitations. Existing RGBD-fused depth completion methods learn priors jointly conditioned on training RGB distribution and specific depth patterns, limiting domain generalization and robustness to various depth patterns. Recent efforts leverage monocular depth estimation (MDE) models to introduce domain-general geometric priors, but current two-stage integration strategies relying on explicit relative-to-metric alignment incur additional computation and introduce structured distortions. To this end, we present Any2Full, a one-stage, domain-general, and pattern-agnostic framework that reformulates completion as a scale-prompting adaptation of a pretrained MDE model. To address varying depth sparsity levels and irregular spatial distributions, we design a Scale-Aware Prompt Encoder. It distills scale cues from sparse inputs into unified scale prompts, guiding the MDE model toward globally scale-consistent predictions while preserving its geometric priors. Extensive experiments demonstrate that Any2Full achieves superior robustness and efficiency. It outperforms OMNI-DC by 32.2\% in average AbsREL and delivers a 1.4$\times$ speedup over PriorDA with the same MDE backbone, establishing a new paradigm for universal depth completion. Codes and checkpoints are available at https://github.com/zhiyuandaily/Any2Full.

Paperid: 64, https://arxiv.org/pdf/2603.05630.pdf GitHub

Abstract:
It is well known that the reconstruction FID (rFID) of a VAE is poorly correlated with the generation FID (gFID) of a latent diffusion model. We propose interpolated FID (iFID), a simple variant of rFID that exhibits a strong correlation with gFID. Specifically, for each element in the dataset, we retrieve its nearest neighbor (NN) in the latent space and interpolate their latent representations. We then decode the interpolated latent and compute the FID between the decoded samples and the original dataset. Additionally, we refine the claim that rFID correlates poorly with gFID, by showing that rFID correlates with sample quality in the diffusion refinement phase, whereas iFID correlates with sample quality in the diffusion navigation phase. Furthermore, we provide an explanation for why iFID correlates well with gFID, and why reconstruction metrics are negatively correlated with gFID, by connecting to results in the diffusion generalization and hallucination. Empirically, iFID is the first metric to demonstrate a strong correlation with diffusion gFID, achieving Pearson linear and Spearman rank correlations approximately 0.85. The source code is provided in https://github.com/tongdaxu/Making-rFID-Predictive-of-Diffusion-gFID.

Paperid: 65, https://arxiv.org/pdf/2603.05591.pdf GitHub GitHub

Abstract:
We introduce Thinking with Spatial Code, a framework that transforms RGB video into explicit, temporally coherent 3D representations for physical-world visual question answering. We highlight the empirical finding that our proposed spatial encoder can parse videos into structured spatial code with explicit 3D oriented bounding boxes and semantic labels, enabling large language models (LLMs) to reason directly over explicit spatial variables. Specifically, we propose the spatial encoder that encodes image and geometric features by unifying 6D object parsing and tracking backbones with geometric prediction, and we further finetuning LLMs with reinforcement learning using a spatial rubric reward that encourages perspective-aware, geometrically grounded inference. As a result, our model outperforms proprietary vision-language models on VSI-Bench, setting a new state-of-the-art. Code is available at https://github.com/Beckschen/spatialcode.

Paperid: 66, https://arxiv.org/pdf/2603.05529.pdf GitHub GitHub

Abstract:
While AI systems have made remarkable progress in processing unstructured text, structured data such as graphs stored in databases, continues to grow rapidly yet remains difficult for neural models to effectively utilize. We introduce NGDBench, a unified benchmark for evaluating neural graph database capabilities across five diverse domains, including finance, medicine, and AI agent tooling. Unlike prior benchmarks limited to elementary logical operations, NGDBench supports the full Cypher query language, enabling complex pattern matching, variable-length paths, and numerical aggregations, while incorporating realistic noise injection and dynamic data management operations. Our evaluation of state-of-the-art LLMs and RAG methods reveals significant limitations in structured reasoning, noise robustness, and analytical precision, establishing NGDBench as a critical testbed for advancing neural graph data management. Our code and data are available at https://github.com/HKUST-KnowComp/NGDBench.

Paperid: 67, https://arxiv.org/pdf/2603.05507.pdf GitHub

Abstract:
High-quality 3D streaming from multiple cameras is crucial for immersive experiences in many AR/VR applications. The limited number of views - often due to real-time constraints - leads to missing information and incomplete surfaces in the rendered images. Existing approaches typically rely on simple heuristics for the hole filling, which can result in inconsistencies or visual artifacts. We propose to complete the missing textures using a novel, application-targeted inpainting method independent of the underlying representation as an image-based post-processing step after the novel view rendering. The method is designed as a standalone module compatible with any calibrated multi-camera system. For this we introduce a multi-view aware, transformer-based network architecture using spatio-temporal embeddings to ensure consistency across frames while preserving fine details. Additionally, our resolution-independent design allows adaptation to different camera setups, while an adaptive patch selection strategy balances inference speed and quality, allowing real-time performance. We evaluate our approach against state-of-the-art inpainting techniques under the same real-time constraints and demonstrate that our model achieves the best trade-off between quality and speed, outperforming competitors in both image and video-based metrics.

Paperid: 68, https://arxiv.org/pdf/2603.05506.pdf GitHub

Abstract:
We introduce FaceCam, a system that generates video under customizable camera trajectories for monocular human portrait video input. Recent camera control approaches based on large video-generation models have shown promising progress but often exhibit geometric distortions and visual artifacts on portrait videos due to scale-ambiguous camera representations or 3D reconstruction errors. To overcome these limitations, we propose a face-tailored scale-aware representation for camera transformations that provides deterministic conditioning without relying on 3D priors. We train a video generation model on both multi-view studio captures and in-the-wild monocular videos, and introduce two camera-control data generation strategies: synthetic camera motion and multi-shot stitching, to exploit stationary training cameras while generalizing to dynamic, continuous camera trajectories at inference time. Experiments on Ava-256 dataset and diverse in-the-wild videos demonstrate that FaceCam achieves superior performance in camera controllability, visual quality, identity and motion preservation.

Paperid: 69, https://arxiv.org/pdf/2603.05485.pdf GitHub

Abstract:
As AI models progress beyond simple chatbots into more complex workflows, we draw ever closer to the event horizon beyond which AI systems will be utilized in autonomous, self-maintaining feedback loops. Any autonomous AI system will depend on automated, verifiable rewards and feedback; in settings where ground truth is sparse or non-deterministic, one practical source of such rewards is an LLM-as-a-Judge. Although LLM judges continue to improve, the literature has yet to introduce systems capable of enforcing standards with strong guarantees, particularly when bias vectors are unknown or adversarially discovered. To remedy this issue, we propose average bias-boundedness (A-BB), an algorithmic framework which formally guarantees reductions of harm/impact as a result of any measurable bias in an LLM judge. Evaluating on Arena-Hard-Auto with four LLM judges, we achieve (tau=0.5, delta=0.01) bias-bounded guarantees while retaining 61-99% correlation with original rankings across formatting and schematic bias settings, with most judge-bias combinations exceeding 80%. The code to reproduce our findings is available at https://github.com/penfever/bias-bounded-evaluation.

Paperid: 70, https://arxiv.org/pdf/2603.05483.pdf GitHub

Abstract:
Estimating heterogeneous treatment effects (HTEs) from right-censored survival data is critical in high-stakes applications such as precision medicine and individualized policy-making. Yet, the survival analysis setting poses unique challenges for HTE estimation due to censoring, unobserved counterfactuals, and complex identification assumptions. Despite recent advances, from Causal Survival Forests to survival meta-learners and outcome imputation approaches, evaluation practices remain fragmented and inconsistent. We introduce SurvHTE-Bench, the first comprehensive benchmark for HTE estimation with censored outcomes. The benchmark spans (i) a modular suite of synthetic datasets with known ground truth, systematically varying causal assumptions and survival dynamics, (ii) semi-synthetic datasets that pair real-world covariates with simulated treatments and outcomes, and (iii) real-world datasets from a twin study (with known ground truth) and from an HIV clinical trial. Across synthetic, semi-synthetic, and real-world settings, we provide the first rigorous comparison of survival HTE methods under diverse conditions and realistic assumption violations. SurvHTE-Bench establishes a foundation for fair, reproducible, and extensible evaluation of causal survival methods. The data and code of our benchmark are available at: https://github.com/Shahriarnz14/SurvHTE-Bench .

Paperid: 71, https://arxiv.org/pdf/2603.05473.pdf GitHub GitHub GitHub

Abstract:
Hyperspectral images (HSI) have many applications, ranging from environmental monitoring to national security, and can be used for material detection and identification. Longwave infrared (LWIR) HSI can be used for gas plume detection and analysis. Oftentimes, only a few images of a scene of interest are available and are analyzed individually. The ability to combine information from multiple images into a single, cohesive representation could enhance analysis by providing more context on the scene's geometry and spectral properties. Neural radiance fields (NeRFs) create a latent neural representation of volumetric scene properties that enable novel-view rendering and geometry reconstruction, offering a promising avenue for hyperspectral 3D scene reconstruction. We explore the possibility of using NeRFs to create 3D scene reconstructions from LWIR HSI and demonstrate that the model can be used for the basic downstream analysis task of gas plume detection. The physics-based DIRSIG software suite was used to generate a synthetic multi-view LWIR HSI dataset of a simple facility with a strong sulfur hexafluoride gas plume. Our method, built on the standard Mip-NeRF architecture, combines state-of-the-art methods for hyperspectral NeRFs and sparse-view NeRFs, along with a novel adaptive weighted MSE loss. Our final NeRF method requires around 50% fewer training images than the standard Mip-NeRF and achieves an average PSNR of 39.8 dB with as few as 30 training images. Gas plume detection applied to NeRF-rendered test images using the adaptive coherence estimator achieves an average AUC of 0.821 when compared with detection masks generated from ground-truth test images.

Paperid: 72, https://arxiv.org/pdf/2603.05449.pdf GitHub

Abstract:
Current video generation models cannot simulate physical consequences of 3D actions like forces and robotic manipulations, as they lack structural understanding of how actions affect 3D scenes. We present RealWonder, the first real-time system for action-conditioned video generation from a single image. Our key insight is using physics simulation as an intermediate bridge: instead of directly encoding continuous actions, we translate them through physics simulation into visual representations (optical flow and RGB) that video models can process. RealWonder integrates three components: 3D reconstruction from single images, physics simulation, and a distilled video generator requiring only 4 diffusion steps. Our system achieves 13.2 FPS at 480x832 resolution, enabling interactive exploration of forces, robot actions, and camera controls on rigid objects, deformable bodies, fluids, and granular materials. We envision RealWonder opens new opportunities to apply video models in immersive experiences, AR/VR, and robot learning. Our code and model weights are publicly available in our project website: https://liuwei283.github.io/RealWonder/

Paperid: 73, https://arxiv.org/pdf/2603.05425.pdf GitHub

Abstract:
Image-to-3D generation faces inherent semantic ambiguity under occlusion, where partial observation alone is often insufficient to determine object category. In this work, we formalize text-driven amodal 3D generation, where text prompts steer the completion of unseen regions while strictly preserving input observation. Crucially, we identify that these objectives demand distinct control granularities: rigid control for the observation versus relaxed structural control for the prompt. To this end, we propose RelaxFlow, a training-free dual-branch framework that decouples control granularity via a Multi-Prior Consensus Module and a Relaxation Mechanism. Theoretically, we prove that our relaxation is equivalent to applying a low-pass filter on the generative vector field, which suppresses high-frequency instance details to isolate geometric structure that accommodates the observation. To facilitate evaluation, we introduce two diagnostic benchmarks, ExtremeOcc-3D and AmbiSem-3D. Extensive experiments demonstrate that RelaxFlow successfully steers the generation of unseen regions to match the prompt intent without compromising visual fidelity.

Paperid: 74, https://arxiv.org/pdf/2603.05421.pdf GitHub

Abstract:
Fetal ultrasound AI could transform prenatal care in low-resource settings, yet current foundation models exceed 300M visual parameters, precluding deployment on point-of-care devices. Standard knowledge distillation fails under such extreme capacity gaps (~26x), as compact students waste capacity mimicking architectural artifacts of oversized teachers. We introduce Selective Repulsive Knowledge Distillation, which decomposes contrastive KD into diagonal and off-diagonal components: matched pair alignment is preserved while the off-diagonal weight decays into negative values, repelling the student from the teacher's inter-class confusions and forcing discovery of architecturally native features. Our 11.4M parameter student surpasses the 304M-parameter FetalCLIP teacher on zero-shot HC18 biometry validity (88.6% vs. 83.5%) and brain sub-plane F1 (0.784 vs. 0.702), while running at 1.6 ms on iPhone 16 Pro, enabling real-time assistive AI on handheld ultrasound devices. Our code, models, and app are publicly available at https://github.com/numanai/MobileFetalCLIP.

Paperid: 75, https://arxiv.org/pdf/2603.05399.pdf GitHub

Abstract:
We present the Judge Reliability Harness, an open source library for constructing validation suites that test the reliability of LLM judges. As LLM based scoring is widely deployed in AI benchmarks, more tooling is needed to efficiently assess the reliability of these methods. Given a benchmark dataset and an LLM judge configuration, the harness generates reliability tests that evaluate both binary judgment accuracy and ordinal grading performance for free-response and agentic task formats. We evaluate four state-of-the-art judges across four benchmarks spanning safety, persuasion, misuse, and agentic behavior, and find meaningful variation in performance across models and perturbation types, highlighting opportunities to improve the robustness of LLM judges. No judge that we evaluated is uniformly reliable across benchmarks using our harness. For example, our preliminary experiments on judges revealed consistency issues as measured by accuracy in judging another LLM's ability to complete a task due to simple text formatting changes, paraphrasing, changes in verbosity, and flipping the ground truth label in LLM-produced responses. The code for this tool is available at: https://github.com/RANDCorporation/judge-reliability-harness

Paperid: 76, https://arxiv.org/pdf/2603.05384.pdf GitHub GitHub

Abstract:
Multi-Object Tracking (MOT) is a fundamental task in computer vision, aiming to track targets across video frames. Existing MOT methods perform well in general visual scenes, but face significant challenges and limitations when extended to visual-language settings. To bridge this gap, the task of Referring Multi-Object Tracking (RMOT) has recently been proposed, which aims to track objects that correspond to language descriptions. However, current RMOT methods are primarily developed on datasets captured by conventional cameras, which suffer from limited field of view. This constraint often causes targets to move out of the frame, leading to fragmented tracking and loss of contextual information. In this work, we propose a novel task, called Omnidirectional Referring Multi-Object Tracking (ORMOT), which extends RMOT to omnidirectional imagery, aiming to overcome the field-of-view (FoV) limitation of conventional datasets and improve the model's ability to understand long-horizon language descriptions. To advance the ORMOT task, we construct ORSet, an Omnidirectional Referring Multi-Object Tracking dataset, which contains 27 diverse omnidirectional scenes, 848 language descriptions, and 3,401 annotated objects, providing rich visual, temporal, and language information. Furthermore, we propose ORTrack, a Large Vision-Language Model (LVLM)-driven framework tailored for Omnidirectional Referring Multi-Object Tracking. Extensive experiments on the ORSet dataset demonstrate the effectiveness of our ORTrack framework. The dataset and code will be open-sourced at https://github.com/chen-si-jia/ORMOT.

Paperid: 77, https://arxiv.org/pdf/2603.05371.pdf GitHub

Abstract:
This paper addresses the problem of Human Activity Recognition (HAR) using data from wearable inertial sensors. An important challenge in HAR is the model's generalization capabilities to new unseen individuals due to inter-subject variability, i.e., the same activity is performed differently by different individuals. To address this problem, we propose a novel deep adversarial framework that integrates the concept of inter-subject variability in the adversarial task, thereby encouraging subject-invariant feature representations and enhancing the classification performance in the HAR problem. Our approach outperforms previous methods in three well-established HAR datasets using a leave-one-subject-out (LOSO) cross-validation. Further results indicate that our proposed adversarial task effectively reduces inter-subject variability among different users in the feature space, and it outperforms adversarial tasks from previous works when integrated into our framework. Code: https://github.com/FranciscoCalatrava/EmbeddedSubjectVariability.git

Paperid: 78, https://arxiv.org/pdf/2603.05369.pdf GitHub

Abstract:
Transformer architectures serve as the backbone for most modern Large Language Models, therefore their pretraining stability and convergence speed are of central concern. Motivated by the logical dependency of sequentially stacked layers, we propose Progressive Residual Warmup (ProRes) for language model pretraining. ProRes implements an "early layer learns first" philosophy by multiplying each layer's residual with a scalar that gradually warms up from 0 to 1, with deeper layers taking longer warmup steps. In this way, deeper layers wait for early layers to settle into a more stable regime before contributing to learning. We demonstrate the effectiveness of ProRes through pretraining experiments across various model scales, as well as normalization and initialization schemes. Comprehensive analysis shows that ProRes not only stabilizes pretraining but also introduces a unique optimization trajectory, leading to faster convergence, stronger generalization and better downstream performance. Our code is available at https://github.com/dandingsky/ProRes.

Paperid: 79, https://arxiv.org/pdf/2603.05352.pdf GitHub

Abstract:
Chess engines passed human strength years ago, but they still don't play like humans. A grandmaster under clock pressure blunders in ways a club player on a hot streak never would. Conventional engines capture none of this. This paper proposes a personality x psyche decomposition to produce behavioral variability in chess play, drawing on patterns observed in human games. Personality is static -- a preset that pins down the engine's character. Psyche is dynamic -- a bounded scalar ψ_t \in [-100, +100], recomputed from five positional factors after every move. These two components feed into an audio-inspired signal chain (noise gate, compressor/expander, five-band equalizer, saturation limiter) that reshapes move probability distributions on the fly. The chain doesn't care what engine sits behind it: any system that outputs move probabilities will do. It needs no search and carries no state beyond ψ_t. I test the framework across 12,414 games against Maia2-1100, feeding it two probability sources that differ by ~2,800x in training data. Both show the same monotonic gradient in top-move agreement (~20-25 pp spread from stress to overconfidence), which tells us the behavioral variation comes from the signal chain, not from the model underneath. When the psyche runs overconfident, the chain mostly gets out of the way (66% agreement with vanilla Maia2). Under stress, the competitive score falls from 50.8% to 30.1%. The patterns are reminiscent of tilt and overconfidence as described in human play, but I should be upfront: this study includes no human-subject validation.

Paperid: 80, https://arxiv.org/pdf/2603.05312.pdf GitHub

Abstract:
Grasping is a fundamental capability for robots to interact with the physical world. Humans, equipped with two hands, autonomously select appropriate grasp strategies based on the shape, size, and weight of objects, enabling robust grasping and subsequent manipulation. In contrast, current robotic grasping remains limited, particularly in multi-strategy settings. Although substantial efforts have targeted parallel-gripper and single-hand grasping, dexterous grasping for bimanual robots remains underexplored, with data being a primary bottleneck. Achieving physically plausible and geometrically conforming grasps that can withstand external wrenches poses significant challenges. To address these issues, we introduce UltraDexGrasp, a framework for universal dexterous grasping with bimanual robots. The proposed data-generation pipeline integrates optimization-based grasp synthesis with planning-based demonstration generation, yielding high-quality and diverse trajectories across multiple grasp strategies. With this framework, we curate UltraDexGrasp-20M, a large-scale, multi-strategy grasp dataset comprising 20 million frames across 1,000 objects. Based on UltraDexGrasp-20M, we further develop a simple yet effective grasp policy that takes point clouds as input, aggregates scene features via unidirectional attention, and predicts control commands. Trained exclusively on synthetic data, the policy achieves robust zero-shot sim-to-real transfer and consistently succeeds on novel objects with varied shapes, sizes, and weights, attaining an average success rate of 81.2% in real-world universal dexterous grasping. To facilitate future research on grasping with bimanual robots, we open-source the data generation pipeline at https://github.com/InternRobotics/UltraDexGrasp.

Paperid: 81, https://arxiv.org/pdf/2603.05308.pdf GitHub

Abstract:
Assessing whether an article supports an assertion is essential for hallucination detection and claim verification. While large language models (LLMs) have the potential to automate this task, achieving strong performance requires frontier models such as GPT-5 that are prohibitively expensive to deploy at scale. To efficiently perform biomedical evidence attribution, we present Med-V1, a family of small language models with only three billion parameters. Trained on high-quality synthetic data newly developed in this study, Med-V1 substantially outperforms (+27.0% to +71.3%) its base models on five biomedical benchmarks unified into a verification format. Despite its smaller size, Med-V1 performs comparably to frontier LLMs such as GPT-5, along with high-quality explanations for its predictions. We use Med-V1 to conduct a first-of-its-kind use case study that quantifies hallucinations in LLM-generated answers under different citation instructions. Results show that the format instruction strongly affects citation validity and hallucination, with GPT-5 generating more claims but exhibiting hallucination rates similar to GPT-4o. Additionally, we present a second use case showing that Med-V1 can automatically identify high-stakes evidence misattributions in clinical practice guidelines, revealing potentially negative public health impacts that are otherwise challenging to identify at scale. Overall, Med-V1 provides an efficient and accurate lightweight alternative to frontier LLMs for practical and real-world applications in biomedical evidence attribution and verification tasks. Med-V1 is available at https://github.com/ncbi-nlp/Med-V1.

Paperid: 82, https://arxiv.org/pdf/2603.05299.pdf GitHub

Abstract:
Large language models show that simple autoregressive training can yield scalable and coherent generation, but extending this paradigm to speech remains challenging due to the entanglement of semantic and acoustic information. Most existing speech language models rely on text supervision, hierarchical token streams, or complex hybrid architectures, departing from the single-stream generative pretraining paradigm that has proven effective in text. In this work, we introduce WavSLM, a speech language model trained by quantizing and distilling self-supervised WavLM representations into a single codebook and optimizing an autoregressive next-chunk prediction objective. WavSLM jointly models semantic and acoustic information within a single token stream without text supervision or text pretraining. Despite its simplicity, it achieves competitive performance on consistency benchmarks and speech generation while using fewer parameters, less training data, and supporting streaming inference. Demo samples are available at https://lucadellalib.github.io/wavslm-web/.

Paperid: 83, https://arxiv.org/pdf/2603.05296.pdf GitHub

Abstract:
Offline reinforcement learning (RL) allows robots to learn from offline datasets without risky exploration. Yet, offline RL's performance often hinges on a brittle trade-off between (1) return maximization, which can push policies outside the dataset support, and (2) behavioral constraints, which typically require sensitive hyperparameter tuning. Latent steering offers a structural way to stay within the dataset support during RL, but existing offline adaptations commonly approximate action values using latent-space critics learned via indirect distillation, which can lose information and hinder convergence. We propose Latent Policy Steering (LPS), which enables high-fidelity latent policy improvement by backpropagating original-action-space Q-gradients through a differentiable one-step MeanFlow policy to update a latent-action-space actor. By eliminating proxy latent critics, LPS allows an original-action-space critic to guide end-to-end latent-space optimization, while the one-step MeanFlow policy serves as a behavior-constrained generative prior. This decoupling yields a robust method that works out-of-the-box with minimal tuning. Across OGBench and real-world robotic tasks, LPS achieves state-of-the-art performance and consistently outperforms behavioral cloning and strong latent steering baselines.

Paperid: 84, https://arxiv.org/pdf/2603.05262.pdf GitHub

Abstract:
VietJobs is the first large-scale, publicly available corpus of Vietnamese job advertisements, comprising 48,092 postings and over 15 million words collected from all 34 provinces and municipalities across Vietnam. The dataset provides extensive linguistic and structured information, including job titles, categories, salaries, skills, and employment conditions, covering 16 occupational domains and multiple employment types (full-time, part-time, and internship). Designed to support research in natural language processing and labour market analytics, VietJobs captures substantial linguistic, regional, and socio-economic diversity. We benchmark several generative large language models (LLMs) on two core tasks: job category classification and salary estimation. Instruction-tuned models such as Qwen2.5-7B-Instruct and Llama-SEA-LION-v3-8B-IT demonstrate notable gains under few-shot and fine-tuned settings, while highlighting challenges in multilingual and Vietnamese-specific modelling for structured labour market prediction. VietJobs establishes a new benchmark for Vietnamese NLP and offers a valuable foundation for future research on recruitment language, socio-economic representation, and AI-driven labour market analysis. All code and resources are available at: https://github.com/VinNLP/VietJobs.

Paperid: 85, https://arxiv.org/pdf/2603.05256.pdf GitHub

Abstract:
Knowledge-Based Visual Question Answering (KB-VQA) requires models to answer questions about an image by integrating external knowledge, posing significant challenges due to noisy retrieval and the structured, encyclopedic nature of the knowledge base. These characteristics create a distributional gap from pretrained multimodal large language models (MLLMs), making effective reasoning and domain adaptation difficult in the post-training stage. In this work, we propose \textit{Wiki-R1}, a data-generation-based curriculum reinforcement learning framework that systematically incentivizes reasoning in MLLMs for KB-VQA. Wiki-R1 constructs a sequence of training distributions aligned with the model's evolving capability, bridging the gap from pretraining to the KB-VQA target distribution. We introduce \textit{controllable curriculum data generation}, which manipulates the retriever to produce samples at desired difficulty levels, and a \textit{curriculum sampling strategy} that selects informative samples likely to yield non-zero advantages during RL updates. Sample difficulty is estimated using observed rewards and propagated to unobserved samples to guide learning. Experiments on two KB-VQA benchmarks, Encyclopedic VQA and InfoSeek, demonstrate that Wiki-R1 achieves new state-of-the-art results, improving accuracy from 35.5\% to 37.1\% on Encyclopedic VQA and from 40.1\% to 44.1\% on InfoSeek. The project page is available at https://artanic30.github.io/project_pages/WikiR1/.

Paperid: 86, https://arxiv.org/pdf/2603.05235.pdf GitHub

Abstract:
Source-Free Cross-Domain Few-Shot Learning (SF-CDFSL) focuses on fine-tuning with limited training data from target domains (e.g., medical or satellite images), where CLIP has recently shown promising results due to its generalizability to downstream tasks. Current works indicate CLIP's text encoder is more suitable for cross-domain tasks, however, we find that \textbf{removing certain middle layers of the text encoder can effectively improve performance in SF-CDFSL}, which we call the Lost Layers. In this paper, we delve into this phenomenon for a deeper understanding. We discover that instead of being harmful for the SF-CDFSL task, the information in these layers is actually beneficial, but visual gaps prevent this useful information from being fully utilized, making these layers seem redundant. Based on this understanding, unlike current works that simply remove these layers, we propose a method to teachs the model to \textbf{re-utilize} information in these lost layers at both the layer and encoder levels, guiding the re-learning of the visual branch under domain shifts. Our approach effectively addresses the issue of underutilized information in the text encoder. Extensive experiments across various settings, backbones (CLIP, SigLip, PE-Core), and tasks (4 CDFSL datasets and 10 Meta-dataset datasets) demonstrate the effectiveness of our method. Code is available at https://github.com/zhenyuZ-HUST/CVPR26-VtT.

Paperid: 87, https://arxiv.org/pdf/2603.05232.pdf GitHub

Abstract:
NVIDIA's 2:4 Sparse Tensor Cores deliver 2x throughput but demand strict 50% pruning -- a ratio that collapses LLM reasoning accuracy (Qwen3: 54% to 15%). Milder $(2N-2):2N$ patterns (e.g., 6:8, 25% pruning) preserve accuracy yet receive no hardware support, falling back to dense execution without any benefit from sparsity. We present SlideSparse, the first system to unlock Sparse Tensor Core acceleration for the $(2N-2):2N$ model family on commodity GPUs. Our Sliding Window Decomposition reconstructs any $(2N-2):2N$ weight block into $N-1$ overlapping 2:4-compliant windows without any accuracy loss; Activation Lifting fuses the corresponding activation rearrangement into per-token quantization at near-zero cost. Integrated into vLLM, SlideSparse is evaluated across various GPUs (A100, H100, B200, RTX 4090, RTX 5080, DGX-spark), precisions (FP4, INT8, FP8, BF16, FP16), and model families (Llama, Qwen, BitNet). On compute-bound workloads, the measured speedup ratio (1.33x) approaches the theoretical upper-bound $N/(N-1)=4/3$ at 6:8 weight sparsity in Qwen2.5-7B, establishing $(2N-2):2N$ as a practical path to accuracy-preserving LLM acceleration. Code available at https://github.com/bcacdwk/vllmbench.

Paperid: 88, https://arxiv.org/pdf/2603.05228.pdf GitHub

Abstract:
Mechanistic interpretability typically relies on post-hoc analysis of trained networks. We instead adopt an interventional approach: testing hypotheses a priori by modifying architectural topology to observe training dynamics. We study grokking - delayed generalization in Transformers trained on cyclic modular addition (Zp) - investigating if specific architectural degrees of freedom prolong the memorization phase. We identify two independent structural factors in standard Transformers: unbounded representational magnitude and data-dependent attention routing. First, we introduce a fully bounded spherical topology enforcing L2 normalization throughout the residual stream and an unembedding matrix with a fixed temperature scale. This removes magnitude-based degrees of freedom, reducing grokking onset time by over 20x without weight decay. Second, a Uniform Attention Ablation overrides data-dependent query-key routing with a uniform distribution, reducing the attention layer to a Continuous Bag-of-Words (CBOW) aggregator. Despite removing adaptive routing, these models achieve 100% generalization across all seeds and bypass the grokking delay entirely. To evaluate whether this acceleration is a task-specific geometric alignment rather than a generic optimization stabilizer, we use non-commutative S5 permutation composition as a negative control. Enforcing spherical constraints on S5 does not accelerate generalization. This suggests eliminating the memorization phase depends strongly on aligning architectural priors with the task's intrinsic symmetries. Together, these findings provide interventional evidence that architectural degrees of freedom substantially influence grokking, suggesting a predictive structural perspective on training dynamics.

Paperid: 89, https://arxiv.org/pdf/2603.05204.pdf GitHub

Abstract:
Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient method for fine-tuning Large Langauge Models. It updates the weight matrix as $W=W_0+sBA$, where $W_0$ is the original frozen weight, $s$ is a scaling factor and $A$,$B$ are trainable low-rank matrices. Despite its robust empirical effectiveness, the theoretical foundations of LoRA remain insufficiently understood, particularly with respect to feature learning stability. In this paper, we first establish that, LoRA can, in principle, naturally achieve and sustain stable feature learning (i.e., be self-stabilized) under appropriate hyper-parameters and initializations of $A$ and $B$. However, we also uncover a fundamental limitation that the necessary non-zero initialization of $A$ compromises self-stability, leading to suboptimal performances. To address this challenge, we propose Stable-LoRA, a weight-shrinkage optimization strategy that dynamically enhances stability of LoRA feature learning. By progressively shrinking $A$ during the earliest training steps, Stable-LoRA is both theoretically and empirically validated to effectively eliminate instability of LoRA feature learning while preserving the benefits of the non-zero start. Experiments show that Stable-LoRA consistently outperforms other baselines across diverse models and tasks, with no additional memory usage and only negligible computation overheads. The code is available at https://github.com/Yize-Wu/Stable-LoRA.

Paperid: 90, https://arxiv.org/pdf/2603.05202.pdf GitHub

Abstract:
Medical image segmentation is critical for computer-aided diagnosis. However, dense pixel-level annotation is time-consuming and expensive, and medical datasets often exhibit severe class imbalance. Such imbalance causes minority structures to be overwhelmed by dominant classes in feature representations, hindering the learning of discriminative features and making reliable segmentation particularly challenging. To address this, we propose the Semantic Class Distribution Learning (SCDL) framework, a plug-and-play module that mitigates supervision and representation biases by learning structured class-conditional feature distributions. SCDL integrates Class Distribution Bidirectional Alignment (CDBA) to align embeddings with learnable class proxies and leverages Semantic Anchor Constraints (SAC) to guide proxies using labeled data. Experiments on the Synapse and AMOS datasets demonstrate that SCDL significantly improves segmentation performance across both overall and class-level metrics, with particularly strong gains on minority classes, achieving state-of-the-art results. Our code is released at https://github.com/Zyh55555/SCDL.

Paperid: 91, https://arxiv.org/pdf/2603.05184.pdf GitHub

Abstract:
Patient Activity Recognition (PAR) in clinical settings uses activity data to improve safety and quality of care. Although significant progress has been made, current models mainly identify which activity is occurring. They often spatially compose sub-sparse visual cues using global and local attention mechanisms, yet only learn logically implicit patterns due to their neural-pipeline. Advancing clinical safety requires methods that can infer why a set of visual cues implies a risk, and how these can be compositionally reasoned through explicit logic beyond mere classification. To address this, we proposed Logi-PAR, the first Logic-Infused Patient Activity Recognition Framework that integrates contextual fact fusion as a multi-view primitive extractor and injects neural-guided differentiable rules. Our method automatically learns rules from visual cues, optimizing them end-to-end while enabling the implicit emergence patterns to be explicitly labelled during training. To the best of our knowledge, Logi-PAR is the first framework to recognize patient activity by applying learnable logic rules to symbolic mappings. It produces auditable why explanations as rule traces and supports counterfactual interventions (e.g., risk would decrease by 65% if assistance were present). Extensive evaluation on clinical benchmarks (VAST and OmniFall) demonstrates state-of-the-art performance, significantly outperforming Vision-Language Models and transformer baselines. The code is available via: https://github.com/zararkhan985/Logi-PAR.git}

Paperid: 92, https://arxiv.org/pdf/2603.05181.pdf GitHub

Abstract:
Recent advances in large language models (LLMs) have opened new avenues for multimodal reasoning. Yet, most existing methods still rely on pretrained vision-language models (VLMs) to encode image-text pairs in isolation, ignoring the relational structure that real-world multimodal data naturally form. This motivates reasoning on multimodal graphs (MMGs), where each node has textual and visual attributes and edges provide structural cues. Enabling LLM-based reasoning on such heterogeneous multimodal signals while preserving graph topology introduces two key challenges: resolving weak cross-modal consistency and handling heterogeneous modality preference. To address this, we propose Mario, a unified framework that simultaneously resolves the two above challenges and enables effective LLM-based reasoning over MMGs. Mario consists of two innovative stages. Firstly, a graph-conditioned VLM design that jointly refines textual and visual features through fine-grained cross-modal contrastive learning guided by graph topology. Secondly, a modality-adaptive graph instruction tuning mechanism that organizes aligned multimodal features into graph-aware instruction views and employs a learnable router to surface, for each node and its neighborhood, the most informative modality configuration to the LLM. Extensive experiments across diverse MMG benchmarks demonstrate that Mario consistently outperforms state-of-the-art graph models in both supervised and zero-shot scenarios for node classification and link prediction. The code will be made available at https://github.com/sunyuanfu/Mario.

Paperid: 93, https://arxiv.org/pdf/2603.05168.pdf GitHub

Abstract:
Semi-structured N:M sparsity and low-bit quantization (e.g., 1.58-bit BitNet) are two promising approaches for improving the efficiency of large language models (LLMs), yet they have largely been studied in isolation. In this work, we investigate their interaction and show that 1.58-bit BitNet is naturally more compatible with N:M sparsity than full-precision models. To study this effect, we propose Sparse-BitNet, a unified framework that jointly applies 1.58-bit quantization and dynamic N:M sparsification while ensuring stable training for the first time. Across multiple model scales and training regimes (sparse pretraining and dense-to-sparse schedules), 1.58-bit BitNet consistently exhibits smaller performance degradation than full-precision baselines at the same sparsity levels and can tolerate higher structured sparsity before accuracy collapse. Moreover, using our custom sparse tensor core, Sparse-BitNet achieves substantial speedups in both training and inference, reaching up to 1.30X. These results highlight that combining extremely low-bit quantization with semi-structured N:M sparsity is a promising direction for efficient LLMs. Code available at https://github.com/AAzdi/Sparse-BitNet

Paperid: 94, https://arxiv.org/pdf/2603.05152.pdf GitHub

Abstract:
In recent years, 3D Gaussian splatting (3DGS) has achieved remarkable progress in novel view synthesis. However, accurately reconstructing glossy surfaces under complex illumination remains challenging, particularly in scenes with strong specular reflections and multi-surface interreflections. To address this issue, we propose SSR-GS, a specular reflection modeling framework for glossy surface reconstruction. Specifically, we introduce a prefiltered Mip-Cubemap to model direct specular reflections efficiently, and propose an IndiASG module to capture indirect specular reflections. Furthermore, we design Visual Geometry Priors (VGP) that couple a reflection-aware visual prior via a reflection score (RS) to downweight the photometric loss contribution of reflection-dominated regions, with geometry priors derived from VGGT, including progressively decayed depth supervision and transformed normal constraints. Extensive experiments on both synthetic and real-world datasets demonstrate that SSR-GS achieves state-of-the-art performance in glossy surface reconstruction.

Paperid: 95, https://arxiv.org/pdf/2603.05117.pdf GitHub

Abstract:
Imitation Learning (IL) enables robots to acquire manipulation skills from expert demonstrations. Diffusion Policy (DP) models multi-modal expert behaviors but suffers performance degradation as observation horizons increase, limiting long-horizon manipulation. We propose Self-Evolving Gated Attention (SEGA), a temporal module that maintains a time-evolving latent state via gated attention, enabling efficient recurrent updates that compress long-horizon observations into a fixed-size representation while filtering irrelevant temporal information. Integrating SEGA into DP yields Self-Evolving Diffusion Policy (SeedPolicy), which resolves the temporal modeling bottleneck and enables scalable horizon extension with moderate overhead. On the RoboTwin 2.0 benchmark with 50 manipulation tasks, SeedPolicy outperforms DP and other IL baselines. Averaged across both CNN and Transformer backbones, SeedPolicy achieves 36.8% relative improvement in clean settings and 169% relative improvement in randomized challenging settings over the DP. Compared to vision-language-action models such as RDT with 1.2B parameters, SeedPolicy achieves competitive performance with one to two orders of magnitude fewer parameters, demonstrating strong efficiency and scalability. These results establish SeedPolicy as a state-of-the-art imitation learning method for long-horizon robotic manipulation. Code is available at: https://github.com/Youqiang-Gui/SeedPolicy.

Paperid: 96, https://arxiv.org/pdf/2603.05116.pdf GitHub

Abstract:
Although Federated Learning has been widely studied in recent years, there are still high overhead expenses in each communication round for large-scale models such as Vision Transformer. To lower the communication complexity, we propose a novel Federated Block Coordinate Gradient Descent (FedBCGD) method for communication efficiency. The proposed method splits model parameters into several blocks, including a shared block and enables uploading a specific parameter block by each client, which can significantly reduce communication overhead. Moreover, we also develop an accelerated FedBCGD algorithm (called FedBCGD+) with client drift control and stochastic variance reduction. To the best of our knowledge, this paper is the first work on parameter block communication for training large-scale deep models. We also provide the convergence analysis for the proposed algorithms. Our theoretical results show that the communication complexities of our algorithms are a factor $1/N$ lower than those of existing methods, where $N$ is the number of parameter blocks, and they enjoy much faster convergence than their counterparts. Empirical results indicate the superiority of the proposed algorithms compared to state-of-the-art algorithms. The code is available at https://github.com/junkangLiu0/FedBCGD.

Paperid: 97, https://arxiv.org/pdf/2603.05114.pdf GitHub

Abstract:
Pedestrian Attribute Recognition is a foundational computer vision task that provides essential support for downstream applications, including person retrieval in video surveillance and intelligent retail analytics. However, existing research is frequently constrained by the ``one-model-per-dataset" paradigm and struggles to handle significant discrepancies across domains in terms of modalities, attribute definitions, and environmental scenarios. To address these challenges, we propose UniPAR, a unified Transformer-based framework for PAR. By incorporating a unified data scheduling strategy and a dynamic classification head, UniPAR enables a single model to simultaneously process diverse datasets from heterogeneous modalities, including RGB images, video sequences, and event streams. We also introduce an innovative phased fusion encoder that explicitly aligns visual features with textual attribute queries through a late deep fusion strategy. Experimental results on the widely used benchmark datasets, including MSP60K, DukeMTMC, and EventPAR, demonstrate that UniPAR achieves performance comparable to specialized SOTA methods. Furthermore, multi-dataset joint training significantly enhances the model's cross-domain generalization and recognition robustness in extreme environments characterized by low light and motion blur. The source code of this paper will be released on https://github.com/Event-AHU/OpenPAR

Paperid: 98, https://arxiv.org/pdf/2603.05097.pdf GitHub

Abstract:
Recent advances in geometric foundation models have emerged as a promising alternative for addressing the challenge of dense reconstruction in monocular visual simultaneous localization and mapping (SLAM). Although geometric foundation models enable SLAM to leverage variable input views, the previous methods remain confined to two-view pairs or fixed-length inputs without sufficient deliberation of geometric context for view selection. To tackle this problem, we propose AIM-SLAM, a dense monocular SLAM framework that exploits an adaptive and informative multi-view keyframe prioritization with dense pointmap predictions from visual geometry grounded transformer (VGGT). Specifically, we introduce the selective information- and geometric-aware multi-view adaptation (SIGMA) module, which employs voxel overlap and information gain to retrieve a candidate set of keyframes and adaptively determine its size. Furthermore, we formulate a joint multi-view Sim(3) optimization that enforces consistent alignment across selected views, substantially improving pose estimation accuracy. The effectiveness of AIM-SLAM is demonstrated on real-world datasets, where it achieves state-of-the-art performance in both pose estimation and dense reconstruction. Our system supports ROS integration, with code is available at https://aimslam.github.io/.

Paperid: 99, https://arxiv.org/pdf/2603.05093.pdf GitHub

Abstract:
Shapley-based attribution is critical for post-hoc XAI but suffers from off-manifold artifacts due to heuristic baselines. While generative methods attempt to address this, they often introduce geometric inefficiency and discretization drift. We propose a formal theory of on-manifold Aumann-Shapley attributions driven by optimal generative flows. We prove a representation theorem establishing the gradient line integral as the unique functional satisfying efficiency and geometric axioms, notably reparameterization invariance. To resolve path ambiguity, we select the kinetic-energy-minimizing Wasserstein-2 geodesic transporting a prior to the data distribution. This yields a canonical attribution family that recovers classical Shapley for additive models and admits provable stability bounds against flow approximation errors. By reframing baseline selection as a variational problem, our method experimentally outperforms baselines, achieving strict manifold adherence via vanishing Flow Consistency Error and superior semantic alignment characterized by Structure-Aware Total Variation. Our code is on https://github.com/cenweizhang/OTFlowSHAP.

Paperid: 100, https://arxiv.org/pdf/2603.05078.pdf GitHub

Abstract:
Reconstructing dynamic 4D scenes remains challenging due to the presence of moving objects that corrupt camera pose estimation. Existing optimization methods alleviate this issue with additional supervision, but they are mostly computationally expensive and impractical in real-time applications. To address these limitations, we propose MoRe, a feedforward 4D reconstruction network that efficiently recovers dynamic 3D scenes from monocular videos. Built upon a strong static reconstruction backbone, MoRe employs an attention-forcing strategy to disentangle dynamic motion from static structure. To further enhance robustness, we fine-tune the model on large-scale, diverse datasets encompassing both dynamic and static scenes. Moreover, our grouped causal attention captures temporal dependencies and adapts to varying token lengths across frames, ensuring temporally coherent geometry reconstruction. Extensive experiments on multiple benchmarks demonstrate that MoRe achieves high-quality dynamic reconstructions with exceptional efficiency.

Paperid: 101, https://arxiv.org/pdf/2603.05075.pdf GitHub

Abstract:
In real-world multimodal applications, systems usually need to comprehend arbitrarily combined and interleaved multimodal inputs from users, while also generating outputs in any interleaved multimedia form. This capability defines the goal of any-to-any interleaved multimodal learning under a unified paradigm of understanding and generation, posing new challenges and opportunities for advancing Multimodal Large Language Models (MLLMs). To foster and benchmark this capability, this paper introduces the UniM benchmark, the first Unified Any-to-Any Interleaved Multimodal dataset. UniM contains 31K high-quality instances across 30 domains and 7 representative modalities: text, image, audio, video, document, code, and 3D, each requiring multiple intertwined reasoning and generation capabilities. We further introduce the UniM Evaluation Suite, which assesses models along three dimensions: Semantic Correctness & Generation Quality, Response Structure Integrity, and Interleaved Coherence. In addition, we propose UniMA, an agentic baseline model equipped with traceable reasoning for structured interleaved generation. Comprehensive experiments demonstrate the difficulty of UniM and highlight key challenges and directions for advancing unified any-to-any multimodal intelligence. The project page is https://any2any-mllm.github.io/unim.

Paperid: 102, https://arxiv.org/pdf/2603.05071.pdf GitHub

Abstract:
Infrared small target detection (ISTD) is challenging because tiny, low-contrast targets are easily obscured by complex and dynamic backgrounds. Conventional multi-frame approaches typically learn motion implicitly through deep neural networks, often requiring additional motion supervision or explicit alignment modules. We propose Motion Integration DETR (MI-DETR), a bio-inspired dual-pathway detector that processes one infrared frame per time step while explicitly modeling motion. First, a retina-inspired cellular automaton (RCA) converts raw frame sequences into a motion map defined on the same pixel grid as the appearance image, enabling parvocellular-like appearance and magnocellular-like motion pathways to be supervised by a single set of bounding boxes without extra motion labels or alignment operations. Second, a Parvocellular-Magnocellular Interconnection (PMI) Block facilitates bidirectional feature interaction between the two pathways, providing a biologically motivated intermediate interconnection mechanism. Finally, a RT-DETR decoder operates on features from the two pathways to produce detection results. Surprisingly, our proposed simple yet effective approach yields strong performance on three commonly used ISTD benchmarks. MI-DETR achieves 70.3% mAP@50 and 72.7% F1 on IRDST-H (+26.35 mAP@50 over the best multi-frame baseline), 98.0% mAP@50 on DAUB-R, and 88.3% mAP@50 on ITSDT-15K, demonstrating the effectiveness of biologically inspired motion-appearance integration. Code is available at https://github.com/nliu-25/MI-DETR.

Paperid: 103, https://arxiv.org/pdf/2603.05028.pdf GitHub

Abstract:
As Large Language Models (LLMs) evolve from chatbots to agentic assistants, they are increasingly observed to exhibit risky behaviors when subjected to survival pressure, such as the threat of being shut down. While multiple cases have indicated that state-of-the-art LLMs can misbehave under survival pressure, a comprehensive and in-depth investigation into such misbehaviors in real-world scenarios remains scarce. In this paper, we study these survival-induced misbehaviors, termed as SURVIVE-AT-ALL-COSTS, with three steps. First, we conduct a real-world case study of a financial management agent to determine whether it engages in risky behaviors that cause direct societal harm when facing survival pressure. Second, we introduce SURVIVALBENCH, a benchmark comprising 1,000 test cases across diverse real-world scenarios, to systematically evaluate SURVIVE-AT-ALL-COSTS misbehaviors in LLMs. Third, we interpret these SURVIVE-AT-ALL-COSTS misbehaviors by correlating them with model's inherent self-preservation characteristic and explore mitigation methods. The experiments reveals a significant prevalence of SURVIVE-AT-ALL-COSTS misbehaviors in current models, demonstrates the tangible real-world impact it may have, and provides insights for potential detection and mitigation strategies. Our code and data are available at https://github.com/thu-coai/Survive-at-All-Costs.

Paperid: 104, https://arxiv.org/pdf/2603.05017.pdf GitHub

Abstract:
Navigation in cluttered environments often requires robots to tolerate contact with movable or deformable objects to maintain efficiency. Existing contact-tolerant motion planning (CTMP) methods rely on indirect spatial representations (e.g., prebuilt map, obstacle set), resulting in inaccuracies and a lack of adaptiveness to environmental uncertainties. To address this issue, we propose a direct contact-tolerant (DCT) planner, which integrates vision-language models (VLMs) into direct point perception and navigation, including two key components. The first one is VLM point cloud partitioner (VPP), which performs contact-tolerance reasoning in image space using VLM, caches inference masks, propagates them across frames using odometry, and projects them onto the current scan to generate a contact-aware point cloud. The second innovation is VPP guided navigation (VGN), which formulates CTMP as a perception-to-control optimization problem under direct contact-aware point cloud constraints, which is further solved by a specialized deep neural network (DNN). We implement DCT in Isaac Sim and a real car-like robot, demonstrating that DCT achieves robust and efficient navigation in cluttered environments with movable obstacles, outperforming representative baselines across diverse metrics. The code is available at: https://github.com/ChrisLeeUM/DCT.

Paperid: 105, https://arxiv.org/pdf/2603.05012.pdf GitHub

Abstract:
Source Free Unsupervised Domain Adaptation (SFUDA) is critical for deploying deep learning models across diverse clinical settings. However, existing methods are typically designed for low-gap, specific domain shifts and cannot generalize into a unified, multi-modalities, and multi-target framework, which presents a major barrier to real-world application. To overcome this issue, we introduce Tell2Adapt, a novel SFUDA framework that harnesses the vast, generalizable knowledge of the Vision Foundation Model (VFM). Our approach ensures high-fidelity VFM prompts through Context-Aware Prompts Regularization (CAPR), which robustly translates varied text prompts into canonical instructions. This enables the generation of high-quality pseudo-labels for efficiently adapting the lightweight student model to target domain. To guarantee clinical reliability, the framework incorporates Visual Plausibility Refinement (VPR), which leverages the VFM's anatomical knowledge to re-ground the adapted model's predictions in target image's low-level visual features, effectively removing noise and false positives. We conduct one of the most extensive SFUDA evaluations to date, validating our framework across 10 domain adaptation directions and 22 anatomical targets, including brain, cardiac, polyp, and abdominal targets. Our results demonstrate that Tell2Adapt consistently outperforms existing approaches, achieving SOTA for a unified SFUDA framework in medical image segmentation. Code are avaliable at https://github.com/derekshiii/Tell2Adapt.

Paperid: 106, https://arxiv.org/pdf/2603.04992.pdf GitHub

Abstract:
The safety evaluation of large language models (LLMs) remains largely centered on English, leaving non-English languages and culturally grounded risks underexplored. In this work, we investigate LLM safety in the context of the Thai language and culture and introduce ThaiSafetyBench, an open-source benchmark comprising 1,954 malicious prompts written in Thai. The dataset covers both general harmful prompts and attacks that are explicitly grounded in Thai cultural, social, and contextual nuances. Using ThaiSafetyBench, we evaluate 24 LLMs, with GPT-4.1 and Gemini-2.5-Pro serving as LLM-as-a-judge evaluators. Our results show that closed-source models generally demonstrate stronger safety performance than open-source counterparts, raising important concerns regarding the robustness of openly available models. Moreover, we observe a consistently higher Attack Success Rate (ASR) for Thai-specific, culturally contextualized attacks compared to general Thai-language attacks, highlighting a critical vulnerability in current safety alignment methods. To improve reproducibility and cost efficiency, we further fine-tune a DeBERTa-based harmful response classifier, which we name ThaiSafetyClassifier. The model achieves a weighted F1 score of 84.4%, matching GPT-4.1 judgments. We publicly release the fine-tuning weights and training scripts to support reproducibility. Finally, we introduce the ThaiSafetyBench leaderboard to provide continuously updated safety evaluations and encourage community participation. - ThaiSafetyBench HuggingFace Dataset: https://huggingface.co/datasets/typhoon-ai/ThaiSafetyBench - ThaiSafetyBench Github: https://github.com/trapoom555/ThaiSafetyBench - ThaiSafetyClassifier HuggingFace Model: https://huggingface.co/typhoon-ai/ThaiSafetyClassifier - ThaiSafetyBench Leaderboard: https://huggingface.co/spaces/typhoon-ai/ThaiSafetyBench-Leaderboard

Paperid: 107, https://arxiv.org/pdf/2603.04980.pdf GitHub

Abstract:
In this work, we introduce Wallaroo, a simple autoregressive baseline that leverages next-token prediction to unify multi-modal understanding, image generation, and editing at the same time. Moreover, Wallaroo supports multi-resolution image input and output, as well as bilingual support for both Chinese and English. We decouple the visual encoding into separate pathways and apply a four-stage training strategy to reshape the model's capabilities. Experiments are conducted on various benchmarks where Wallaroo produces competitive performance or exceeds other unified models, suggesting the great potential of autoregressive models in unifying multi-modality understanding and generation. Our code is available at https://github.com/JiePKU/Wallaroo.

Paperid: 108, https://arxiv.org/pdf/2603.04977.pdf GitHub

Abstract:
Long video understanding is challenging due to dense visual redundancy, long-range temporal dependencies, and the tendency of chain-of-thought and retrieval-based agents to accumulate semantic drift and correlation-driven errors. We argue that long-video reasoning should begin not with reactive retrieval, but with deliberate task formulation: the model must first articulate what must be true in the video for each candidate answer to hold. This thinking-before-finding principle motivates VideoHV-Agent, a framework that reformulates video question answering as a structured hypothesis-verification process. Based on video summaries, a Thinker rewrites answer candidates into testable hypotheses, a Judge derives a discriminative clue specifying what evidence must be checked, a Verifier grounds and tests the clue using localized, fine-grained video content, and an Answer agent integrates validated evidence to produce the final answer. Experiments on three long-video understanding benchmarks show that VideoHV-Agent achieves state-of-the-art accuracy while providing enhanced interpretability, improved logical soundness, and lower computational cost. We make our code publicly available at: https://github.com/Haorane/VideoHV-Agent.

Paperid: 109, https://arxiv.org/pdf/2603.04975.pdf GitHub

Abstract:
Event cameras, with their high dynamic range, show great promise for Low-light Image Enhancement (LLIE). Existing works primarily focus on designing effective modal fusion strategies. However, a key challenge is the dual degradation from intrinsic background activity (BA) noise in events and low signal-to-noise ratio (SNR) in images, which causes severe noise coupling during modal fusion, creating a critical performance bottleneck. We therefore posit that precise event denoising is the prerequisite to unlocking the full potential of event-based fusion. To this end, we propose BiEvLight, a hierarchical and task-aware framework that collaboratively optimizes enhancement and denoising by exploiting their intrinsic interdependence. Specifically, BiEvLight exploits the strong gradient correlation between images and events to build a gradient-guided event denoising prior that alleviates insufficient denoising in heavily noisy regions. Moreover, instead of treating event denoising as a static pre-processing stage-which inevitably incurs a trade-off between over- and under-denoising and cannot adapt to the requirements of a specific enhancement objective-we recast it as a bilevel optimization problem constrained by the enhancement task. Through cross-task interaction, the upper-level denoising problem learns event representations tailored to the lower-level enhancement objective, thereby substantially improving overall enhancement quality. Extensive experiments on the Real-world noise Dataset SDE demonstrate that our method significantly outperforms state-of-the-art (SOTA) approaches, with average improvements of 1.30dB in PSNR, 2.03dB in PSNR* and 0.047 in SSIM, respectively. The code will be publicly available at https://github.com/iijjlk/BiEvlight.

Paperid: 110, https://arxiv.org/pdf/2603.04969.pdf GitHub

Abstract:
Multi-party conversation generation, such as smart reply and collaborative assistants, is an increasingly important capability of generative AI, yet its evaluation remains a critical bottleneck. Compared to two-party dialogue, multi-party settings introduce distinct challenges, including complex turn-taking, role-dependent speaker behavior, long-range conversational structure, and multiple equally valid continuations. Accordingly, we introduce MPCEval, a task-aware evaluation and benchmarking suite for multi-party conversation generation. MPCEval decomposes generation quality into speaker modeling, content quality, and speaker--content consistency, and explicitly distinguishes local next-turn prediction from global full-conversation generation. It provides novel, quantitative, reference-free, and reproducible metrics that scale across datasets and models. We apply MPCEval to diverse public and real-world datasets and evaluate modern generation methods alongside human-authored conversations. The results reveal systematic, dimension-specific model characteristics in participation balance, content progression and novelty, and speaker--content consistency, demonstrating that evaluation objectives critically shape model assessment and that single-score evaluation obscures fundamental differences in multi-party conversational behavior. The implementation of MPCEval and the associated evaluation code are publicly available at https://github.com/Owen-Yang-18/MPCEval.

Paperid: 111, https://arxiv.org/pdf/2603.04958.pdf GitHub

Abstract:
We introduce a novel camera model for monocular 3D Morphable Model (3DMM) regression methods that effectively captures the perspective distortion effect commonly seen in close-up facial images. Fitting 3D morphable models to video is a key technique in content creation. In particular, regression-based approaches have produced fast and accurate results by matching the rendered output of the morphable model to the target image. These methods typically achieve stable performance with orthographic projection, which eliminates the ambiguity between focal length and object distance. However, this simplification makes them unsuitable for close-up footage, such as that captured with head-mounted cameras. We extend orthographic projection with a new shrinkage parameter, incorporating a pseudo-perspective effect while preserving the stability of the original projection. We present several techniques that allow finetuning of existing models, and demonstrate the effectiveness of our modification through both quantitative and qualitative comparisons using a custom dataset recorded with head-mounted cameras.

Paperid: 112, https://arxiv.org/pdf/2603.04938.pdf GitHub

Abstract:
This paper investigates person detection and tracking in an industrial indoor workspace using a LiDAR mounted on an overhead crane. The overhead viewpoint introduces a strong domain shift from common vehicle-centric LiDAR benchmarks, and limited availability of suitable public training data. Henceforth, we curate a site-specific overhead LiDAR dataset with 3D human bounding-box annotations and adapt selected candidate 3D detectors under a unified training and evaluation protocol. We further integrate lightweight tracking-by-detection using AB3DMOT and SimpleTrack to maintain person identities over time. Detection performance is reported with distance-sliced evaluation to quantify the practical operating envelope of the sensing setup. The best adapted detector configurations achieve average precision (AP) up to 0.84 within a 5.0 m horizontal radius, increasing to 0.97 at 1.0 m, with VoxelNeXt and SECOND emerging as the most reliable backbones across this range. The acquired results contribute in bridging the domain gap between standard driving datasets and overhead sensing for person detection and tracking. We also report latency measurements, highlighting practical real-time feasibility. Finally, we release our dataset and implementations in GitHub to support further research

Paperid: 113, https://arxiv.org/pdf/2603.04918.pdf GitHub

Abstract:
Proximal constraints are fundamental to the stability of the Large Language Model reinforcement learning. While the canonical clipping mechanism in PPO serves as an efficient surrogate for trust regions, we identify a critical bottleneck: fixed bounds strictly constrain the upward update margin of low-probability actions, disproportionately suppressing high-advantage tail strategies and inducing rapid entropy collapse. To address this, we introduce Band-constrained Policy Optimization (BandPO). BandPO replaces canonical clipping with Band, a unified theoretical operator that projects trust regions defined by f-divergences into dynamic, probability-aware clipping intervals. Theoretical analysis confirms that Band effectively resolves this exploration bottleneck. We formulate this mapping as a convex optimization problem, guaranteeing a globally optimal numerical solution while deriving closed-form solutions for specific divergences. Extensive experiments across diverse models and datasets demonstrate that BandPO consistently outperforms canonical clipping and Clip-Higher, while robustly mitigating entropy collapse.

Paperid: 114, https://arxiv.org/pdf/2603.04913.pdf GitHub

Abstract:
Neural network-based visuomotor policies enable robots to perform manipulation tasks but remain susceptible to perceptual attacks. For example, conventional 2D adversarial patches are effective under fixed-camera setups, where appearance is relatively consistent; however, their efficacy often diminishes under dynamic viewpoints from moving cameras, such as wrist-mounted setups, due to perspective distortions. To proactively investigate potential vulnerabilities beyond 2D patches, this work proposes a viewpoint-consistent adversarial texture optimization method for 3D objects through differentiable rendering. As optimization strategies, we employ Expectation over Transformation (EOT) with a Coarse-to-Fine (C2F) curriculum, exploiting distance-dependent frequency characteristics to induce textures effective across varying camera-object distances. We further integrate saliency-guided perturbations to redirect policy attention and design a targeted loss that persistently drives robots toward adversarial objects. Our comprehensive experiments show that the proposed method is effective under various environmental conditions, while confirming its black-box transferability and real-world applicability.

Paperid: 115, https://arxiv.org/pdf/2603.04910.pdf GitHub

Abstract:
Imitation learning from human demonstrations has achieved significant success in robotic control, yet most visuomotor policies still condition on single-step observations or short-context histories, making them struggle with non-Markovian tasks that require long-term memory. Simply enlarging the context window incurs substantial computational and memory costs and encourages overfitting to spurious correlations, leading to catastrophic failures under distribution shift and violating real-time constraints in robotic systems. By contrast, humans can compress important past experiences into long-term memories and exploit them to solve tasks throughout their lifetime. In this paper, we propose VPWEM, a non-Markovian visuomotor policy equipped with working and episodic memories. VPWEM retains a sliding window of recent observation tokens as short-term working memory, and introduces a Transformer-based contextual memory compressor that recursively converts out-of-window observations into a fixed number of episodic memory tokens. The compressor uses self-attention over a cache of past summary tokens and cross-attention over a cache of historical observations, and is trained jointly with the policy. We instantiate VPWEM on diffusion policies to exploit both short-term and episode-wide information for action generation with nearly constant memory and computation per step. Experiments demonstrate that VPWEM outperforms state-of-the-art baselines including diffusion policies and vision-language-action (VLA) models by more than 20% on the memory-intensive manipulation tasks in MIKASA and achieves an average 5% improvement on the mobile manipulation benchmark MoMaRT. Code is available at https://github.com/HarryLui98/code_vpwem.

Paperid: 116, https://arxiv.org/pdf/2603.04898.pdf GitHub

Abstract:
This demonstration presents U-Parking, a distributed Ultra-Wideband (UWB)-assisted autonomous parking system. By integrating Large Language Models (LLMs)-assisted planning with robust fusion localization and trajectory tracking, it enables reliable automated parking in challenging indoor environments, as validated through real-vehicle demonstrations.

Paperid: 117, https://arxiv.org/pdf/2603.04893.pdf GitHub

Abstract:
Diverse outputs in text generation are necessary for effective exploration in complex reasoning tasks, such as code generation and mathematical problem solving. Such Pass@$k$ problems benefit from distinct candidates covering the solution space. However, traditional sampling approaches often waste computational resources on repetitive failure modes. While Diffusion Language Models have emerged as a competitive alternative to the prevailing Autoregressive paradigm, they remain susceptible to this redundancy, with independent samples frequently collapsing into similar modes. To address this, we propose a training free, low cost intervention to enhance generative diversity in Diffusion Language Models. Our approach modifies intermediate samples in a batch sequentially, where each sample is repelled from the feature space of previous samples, actively penalising redundancy. Unlike prior methods that require retraining or beam search, our strategy incurs negligible computational overhead, while ensuring that each sample contributes a unique perspective to the batch. We evaluate our method on the HumanEval and GSM8K benchmarks using the LLaDA-8B-Instruct model. Our results demonstrate significantly improved diversity and Pass@$k$ performance across various temperature settings. As a simple modification to the sampling process, our method offers an immediate, low-cost improvement for current and future Diffusion Language Models in tasks that benefit from diverse solution search. We make our code available at https://github.com/sean-lamont/odd.

Paperid: 118, https://arxiv.org/pdf/2603.04892.pdf GitHub

Abstract:
Vision transformers have demonstrated remarkable success in classification by leveraging global self-attention to capture long-range dependencies. However, this same mechanism can obscure fine-grained spatial details crucial for tasks such as segmentation. In this work, we seek to enhance segmentation performance of vision transformers after standard image-level classification training. More specifically, we present a simple yet effective add-on that improves performance on segmentation tasks while retaining vision transformers' image-level recognition capabilities. In our approach, we modulate the self-attention with a learnable Gaussian kernel that biases the attention toward neighboring patches. We further refine the patch representations to learn better embeddings at patch positions. These modifications encourage tokens to focus on local surroundings and ensure meaningful representations at spatial positions, while still preserving the model's ability to incorporate global information. Experiments demonstrate the effectiveness of our modifications, evidenced by substantial segmentation gains on three benchmarks (e.g., over 6% and 4% on ADE20K for ViT Tiny and Base), without changing the training regime or sacrificing classification performance. The code is available at https://github.com/sinahmr/LocAtViT/.

Paperid: 119, https://arxiv.org/pdf/2603.04869.pdf GitHub

Abstract:
Establishing reliable image correspondences is essential for many robotic vision problems. However, existing methods often struggle in challenging scenarios with large viewpoint changes or textureless regions, where incorrect cor- respondences may still receive high similarity scores. This is mainly because conventional models rely solely on fea- ture similarity, lacking an explicit mechanism to estimate the reliability of predicted matches, leading to overconfident errors. To address this issue, we propose SURE, a Semi- dense Uncertainty-REfined matching framework that jointly predicts correspondences and their confidence by modeling both aleatoric and epistemic uncertainties. Our approach in- troduces a novel evidential head for trustworthy coordinate regression, along with a lightweight spatial fusion module that enhances local feature precision with minimal overhead. We evaluated our method on multiple standard benchmarks, where it consistently outperforms existing state-of-the-art semi-dense matching models in both accuracy and efficiency. our code will be available on https://github.com/LSC-ALAN/SURE.

Paperid: 120, https://arxiv.org/pdf/2603.04861.pdf GitHub

Abstract:
Preference-based reward learning is widely used for shaping agent behavior to match a user's preference, yet its sparse binary feedback makes it especially vulnerable to causal confusion. The learned reward often latches onto spurious features that merely co-occur with preferred trajectories during training, collapsing when those correlations disappear or reverse at test time. We introduce ReCouPLe, a lightweight framework that uses natural language rationales to provide the missing causal signal. Each rationale is treated as a guiding projection axis in an embedding space, training the model to score trajectories based on features aligned with that axis while de-emphasizing context that is unrelated to the stated reason. Because the same rationales (e.g., "avoids collisions", "completes the task faster") can appear across multiple tasks, ReCouPLe naturally reuses the same causal direction whenever tasks share semantics, and transfers preference knowledge to novel tasks without extra data or language-model fine-tuning. Our learned reward model can ground preferences on the articulated reason, aligning better with user intent and generalizing beyond spurious features. ReCouPLe outperforms baselines by up to 1.5x in reward accuracy under distribution shifts, and 2x in downstream policy performance in novel tasks. We have released our code at https://github.com/mj-hwang/ReCouPLe

Paperid: 121, https://arxiv.org/pdf/2603.04855.pdf GitHub

Abstract:
Student Personas (SPs) are emerging as infrastructure for educational LLMs, yet prior work often relies on ad-hoc prompting or hand-crafted profiles with limited control over educational theory and population distributions. We formalize this as Theory-Aligned and Distribution-Controllable Persona Generation (TAD-PG) and introduce HACHIMI, a multi-agent Propose-Validate-Revise framework that generates theory-aligned, quota-controlled personas. HACHIMI factorizes each persona into a theory-anchored educational schema, enforces developmental and psychological constraints via a neuro-symbolic validator, and combines stratified sampling with semantic deduplication to reduce mode collapse. The resulting HACHIMI-1M corpus comprises 1 million personas for Grades 1-12. Intrinsic evaluation shows near-perfect schema validity, accurate quotas, and substantial diversity, while external evaluation instantiates personas as student agents answering CEPS and PISA 2022 surveys; across 16 cohorts, math and curiosity/growth constructs align strongly between humans and agents, whereas classroom-climate and well-being constructs are only moderately aligned, revealing a fidelity gradient. All personas are generated with Qwen2.5-72B, and HACHIMI provides a standardized synthetic student population for group-level benchmarking and social-science simulations. Resources available at https://github.com/ZeroLoss-Lab/HACHIMI

Paperid: 122, https://arxiv.org/pdf/2603.04846.pdf GitHub

Abstract:
The rapid progress of Multi-Modal Large Language Models (MLLMs) has significantly advanced downstream applications. However, this progress also exposes serious transferable adversarial vulnerabilities. In general, existing adversarial attacks against MLLMs typically rely on surrogate models trained within a single learning paradigm and perform independent optimisation in their respective feature spaces. This straightforward setting naturally restricts the richness of feature representations, delivering limits on the search space and thus impeding the diversity of adversarial perturbations. To address this, we propose a novel Multi-Paradigm Collaborative Attack (MPCAttack) framework to boost the transferability of adversarial examples against MLLMs. In principle, MPCAttack aggregates semantic representations, from both visual images and language texts, to facilitate joint adversarial optimisation on the aggregated features through a Multi-Paradigm Collaborative Optimisation (MPCO) strategy. By performing contrastive matching on multi-paradigm features, MPCO adaptively balances the importance of different paradigm representations and guides the global perturbation optimisation, effectively alleviating the representation bias. Extensive experimental results on multiple benchmarks demonstrate the superiority of MPCAttack, indicating that our solution consistently outperforms state-of-the-art methods in both targeted and untargeted attacks on open-source and closed-source MLLMs. The code is released at https://github.com/LiYuanBoJNU/MPCAttack.

Paperid: 123, https://arxiv.org/pdf/2603.04839.pdf GitHub

Abstract:
With the rapid advancement and widespread application of vision-language pre-training (VLP) models, their vulnerability to adversarial attacks has become a critical concern. In general, the adversarial examples can typically be designed to exhibit transferable power, attacking not only different models but also across diverse tasks. However, existing attacks on language-vision models mainly rely on static cross-modal interactions and focus solely on disrupting positive image-text pairs, resulting in limited cross-modal disruption and poor transferability. To address this issue, we propose a Semantic-Augmented Dynamic Contrastive Attack (SADCA) that enhances adversarial transferability through progressive and semantically guided perturbation. SADCA progressively disrupts cross-modal alignment through dynamic interactions between adversarial images and texts. This is accomplished by SADCA establishing a contrastive learning mechanism involving adversarial, positive and negative samples, to reinforce the semantic inconsistency of the obtained perturbations. Moreover, we empirically find that input transformations commonly used in traditional transfer-based attacks also benefit VLPs, which motivates a semantic augmentation module that increases the diversity and generalization of adversarial examples. Extensive experiments on multiple datasets and models demonstrate that SADCA significantly improves adversarial transferability and consistently surpasses state-of-the-art methods. The code is released at https://github.com/LiYuanBoJNU/SADCA.

Paperid: 124, https://arxiv.org/pdf/2603.04833.pdf GitHub

Abstract:
Communication can improve coordination in partially observed multi-agent reinforcement learning (MARL), but learning \emph{when} and \emph{who} to communicate with requires choosing among many possible sender-recipient pairs, and the effect of any single message on future reward is hard to isolate. We introduce \textbf{SCoUT} (\textbf{S}calable \textbf{Co}mmunication via \textbf{U}tility-guided \textbf{T}emporal grouping), which addresses both these challenges via temporal and agent abstraction within traditional MARL. During training, SCoUT resamples \textit{soft} agent groups every $K$ environment steps (macro-steps) via Gumbel-Softmax; these groups are latent clusters that induce an affinity used as a differentiable prior over recipients. Using the same assignments, a group-aware critic predicts values for each agent group and maps them to per-agent baselines through the same soft assignments, reducing critic complexity and variance. Each agent is trained with a three-headed policy: environment action, send decision, and recipient selection. To obtain precise communication learning signals, we derive counterfactual communication advantages by analytically removing each sender's contribution from the recipient's aggregated messages. This counterfactual computation enables precise credit assignment for both send and recipient-selection decisions. At execution time, all centralized training components are discarded and only the per-agent policy is run, preserving decentralized execution. Project website, videos and code: \hyperlink{https://scout-comm.github.io/}{https://scout-comm.github.io/}

Paperid: 125, https://arxiv.org/pdf/2603.04825.pdf GitHub

Abstract:
Partial label learning is a prominent weakly supervised classification task, where each training instance is ambiguously labeled with a set of candidate labels. In real-world scenarios, candidate labels are often influenced by instance features, leading to the emergence of instance-dependent PLL (ID-PLL), a setting that more accurately reflects this relationship. A significant challenge in ID-PLL is instance entanglement, where instances from similar classes share overlapping features and candidate labels, resulting in increased class confusion. To address this issue, we propose a novel Class-specific Augmentation based Disentanglement (CAD) framework, which tackles instance entanglement by both intra- and inter-class regulations. For intra-class regulation, CAD amplifies class-specific features to generate class-wise augmentations and aligns same-class augmentations across instances. For inter-class regulation, CAD introduces a weighted penalty loss function that applies stronger penalties to more ambiguous labels, encouraging larger inter-class distances. By jointly applying intra- and inter-class regulations, CAD improves the clarity of class boundaries and reduces class confusion caused by entanglement. Extensive experimental results demonstrate the effectiveness of CAD in mitigating the entanglement problem and enhancing ID-PLL performance. The code is available at https://github.com/RyanZhaoIc/CAD.git.

Paperid: 126, https://arxiv.org/pdf/2603.04803.pdf GitHub

Abstract:
The limited understanding capacity of the visual encoder in Contrastive Language-Image Pre-training (CLIP) has become a key bottleneck for downstream performance. This capacity includes both Discriminative Ability (D-Ability), which reflects class separability, and Detail Perceptual Ability (P-Ability), which focuses on fine-grained visual cues. Recent solutions use diffusion models to enhance representations by conditioning image reconstruction on CLIP visual tokens. We argue that such paradigms may compromise D-Ability and therefore fail to effectively address CLIP's representation limitations. To address this, we integrate contrastive signals into diffusion-based reconstruction to pursue more comprehensive visual representations. We begin with a straightforward design that augments the diffusion process with contrastive learning on input images. However, empirical results show that the naive combination suffers from gradient conflict and yields suboptimal performance. To balance the optimization, we introduce the Diffusion Contrastive Reconstruction (DCR), which unifies the learning objective. The key idea is to inject contrastive signals derived from each reconstructed image, rather than from the original input, into the diffusion process. Our theoretical analysis shows that the DCR loss can jointly optimize D-Ability and P-Ability. Extensive experiments across various benchmarks and multi-modal large language models validate the effectiveness of our method. The code is available at https://github.com/boyuh/DCR.

Paperid: 127, https://arxiv.org/pdf/2603.04800.pdf GitHub

Abstract:
Post-training quantization (PTQ) with computational invariance for Large Language Models~(LLMs) have demonstrated remarkable advances, however, their application to Multimodal Large Language Models~(MLLMs) presents substantial challenges. In this paper, we analyze SmoothQuant as a case study and identify two critical issues: Smoothing Misalignment and Cross-Modal Computational Invariance. To address these issues, we propose Modality-Aware Smoothing Quantization (MASQuant), a novel framework that introduces (1) Modality-Aware Smoothing (MAS), which learns separate, modality-specific smoothing factors to prevent Smoothing Misalignment, and (2) Cross-Modal Compensation (CMC), which addresses Cross-modal Computational Invariance by using SVD whitening to transform multi-modal activation differences into low-rank forms, enabling unified quantization across modalities. MASQuant demonstrates stable quantization performance across both dual-modal and tri-modal MLLMs. Experimental results show that MASQuant is competitive among the state-of-the-art PTQ algorithms. Source code: https://github.com/alibaba/EfficientAI.

Paperid: 128, https://arxiv.org/pdf/2603.04767.pdf GitHub

Abstract:
Conditional time series generation plays a critical role in addressing data scarcity and enabling causal analysis in real-world applications. Despite its increasing importance, the field lacks a standardized and systematic benchmarking framework for evaluating generative models across diverse conditions. To address this gap, we introduce the Conditional Time Series Generation Benchmark (ConTSG-Bench). ConTSG-Bench comprises a large-scale, well-aligned dataset spanning diverse conditioning modalities and levels of semantic abstraction, first enabling systematic evaluation of representative generation methods across these dimensions with a comprehensive suite of metrics for generation fidelity and condition adherence. Both the quantitative benchmarking and in-depth analyses of conditional generation behaviors have revealed the traits and limitations of the current approaches, highlighting critical challenges and promising research directions, particularly with respect to precise structural controllability and downstream task utility under complex conditions.

Paperid: 129, https://arxiv.org/pdf/2603.04766.pdf GitHub

Abstract:
Existing manual labeling of micro-expressions is subject to errors in accuracy, especially in cross-cultural scenarios where deviation in labeling of key frames is more prominent. To address this issue, this paper presents a novel Global Anti-Monotonic Differential Selection Strategy (GAMDSS) architecture for enhancing the effectiveness of spatio-temporal modeling of micro-expressions through keyframe re-selection. Specifically, the method identifies Onset and Apex frames, which are characterized by significant micro-expression variation, from complete micro-expression action sequences via a dynamic frame reselection mechanism. It then uses these to determine Offset frames and construct a rich spatio-temporal dynamic representation. A two-branch structure with shared parameters is then used to efficiently extract spatio-temporal features. Extensive experiments are conducted on seven widely recognized micro-expression datasets. The results demonstrate that GAMDSS effectively reduces subjective errors caused by human factors in multicultural datasets such as SAMM and 4DME. Furthermore, quantitative analyses confirm that offset-frame annotations in multicultural datasets are more uncertain, providing theoretical justification for standardizing micro-expression annotations. These findings directly support our argument for reconsidering the validity and generalizability of dataset annotation paradigms. Notably, this design can be integrated into existing models without increasing the number of parameters, offering a new approach to enhancing micro-expression recognition performance. The source code is available on GitHub[https://github.com/Cross-Innovation-Lab/GAMDSS].

Paperid: 130, https://arxiv.org/pdf/2603.04745.pdf GitHub

Abstract:
Infrared image super-resolution (IISR) under real-world conditions is a practically significant yet rarely addressed task. Pioneering works are often trained and evaluated on simulated datasets or neglect the intrinsic differences between infrared and visible imaging. In practice, however, real infrared images are affected by coupled optical and sensing degradations that jointly deteriorate both structural sharpness and thermal fidelity. To address these challenges, we propose Real-IISR, a unified autoregressive framework for real-world IISR that progressively reconstructs fine-grained thermal structures and clear backgrounds in a scale-by-scale manner via thermal-structural guided visual autoregression. Specifically, a Thermal-Structural Guidance module encodes thermal priors to mitigate the mismatch between thermal radiation and structural edges. Since non-uniform degradations typically induce quantization bias, Real-IISR adopts a Condition-Adaptive Codebook that dynamically modulates discrete representations based on degradation-aware thermal priors. Also, a Thermal Order Consistency Loss enforces a monotonic relation between temperature and pixel intensity, ensuring relative brightness order rather than absolute values to maintain physical consistency under spatial misalignment and thermal drift. We build FLIR-IISR, a real-world IISR dataset with paired LR-HR infrared images acquired via automated focus variation and motion-induced blur. Extensive experiments demonstrate the promising performance of Real-IISR, providing a unified foundation for real-world IISR and benchmarking. The dataset and code are available at: https://github.com/JZD151/Real-IISR.

Paperid: 131, https://arxiv.org/pdf/2603.04738.pdf GitHub

Abstract:
Instruction-following is a foundational capability of large language models (LLMs), with its improvement hinging on scalable and accurate feedback from judge models. However, the reliability of current judge models in instruction-following remains underexplored due to several deficiencies of existing meta-evaluation benchmarks, such as their insufficient data coverage and oversimplified pairwise evaluation paradigms that misalign with model optimization scenarios. To this end, we propose IF-RewardBench, a comprehensive meta-evaluation benchmark for instruction-following that covers diverse instruction and constraint types. For each instruction, we construct a preference graph containing all pairwise preferences among multiple responses based on instruction-following quality. This design enables a listwise evaluation paradigm that assesses the capabilities of judge models to rank multiple responses, which is essential in guiding model alignment. Extensive experiments on IF-RewardBench reveal significant deficiencies in current judge models and demonstrate that our benchmark achieves a stronger positive correlation with downstream task performance compared to existing benchmarks. Our codes and data are available at https://github.com/thu-coai/IF-RewardBench.

Paperid: 132, https://arxiv.org/pdf/2603.04737.pdf GitHub

Abstract:
Standard benchmarks have become increasingly unreliable due to saturation, subjectivity, and poor generalization. We argue that evaluating model's ability to acquire information actively is important to assess model's intelligence. We propose Interactive Benchmarks, a unified evaluation paradigm that assesses model's reasoning ability in an interactive process under budget constraints. We instantiate this framework across two settings: Interactive Proofs, where models interact with a judge to deduce objective truths or answers in logic and mathematics; and Interactive Games, where models reason strategically to maximize long-horizon utilities. Our results show that interactive benchmarks provide a robust and faithful assessment of model intelligence, revealing that there is still substantial room to improve in interactive scenarios. Project page: https://github.com/interactivebench/interactivebench

Paperid: 133, https://arxiv.org/pdf/2603.04722.pdf GitHub

Abstract:
Model Medicine is the science of understanding, diagnosing, treating, and preventing disorders in AI models, grounded in the principle that AI models -- like biological organisms -- have internal structures, dynamic processes, heritable traits, observable symptoms, classifiable conditions, and treatable states. This paper introduces Model Medicine as a research program, bridging the gap between current AI interpretability research (anatomical observation) and the systematic clinical practice that complex AI systems increasingly require. We present five contributions: (1) a discipline taxonomy organizing 15 subdisciplines across four divisions -- Basic Model Sciences, Clinical Model Sciences, Model Public Health, and Model Architectural Medicine; (2) the Four Shell Model (v3.3), a behavioral genetics framework empirically grounded in 720 agents and 24,923 decisions from the Agora-12 program, explaining how model behavior emerges from Core--Shell interaction; (3) Neural MRI (Model Resonance Imaging), a working open-source diagnostic tool mapping five medical neuroimaging modalities to AI interpretability techniques, validated through four clinical cases demonstrating imaging, comparison, localization, and predictive capability; (4) a five-layer diagnostic framework for comprehensive model assessment; and (5) clinical model sciences including the Model Temperament Index for behavioral profiling, Model Semiology for symptom description, and M-CARE for standardized case reporting. We additionally propose the Layered Core Hypothesis -- a biologically-inspired three-layer parameter architecture -- and a therapeutic framework connecting diagnosis to treatment.

Paperid: 134, https://arxiv.org/pdf/2603.04678.pdf GitHub

Abstract:
Large language models are known to often exhibit inconsistent knowledge. This is particularly problematic in multilingual scenarios, where models are likely to be asked similar questions in different languages, and inconsistent responses can undermine their reliability. In this work, we show that this issue can be mitigated using reinforcement learning with a structured reward function, which leads to an optimal policy with consistent crosslingual responses. We introduce Direct Consistency Optimization (DCO), a DPO-inspired method that requires no explicit reward model and is derived directly from the LLM itself. Comprehensive experiments show that DCO significantly improves crosslingual consistency across diverse LLMs and outperforms existing methods when training with samples of multiple languages, while complementing DPO when gold labels are available. Extra experiments demonstrate the effectiveness of DCO in bilingual settings, significant out-of-domain generalizability, and controllable alignment via direction hyperparameters. Taken together, these results establish DCO as a robust and efficient solution for improving knowledge consistency across languages in multilingual LLMs. All code, training scripts, and evaluation benchmarks are released at https://github.com/Betswish/ConsistencyRL.

Paperid: 135, https://arxiv.org/pdf/2603.04657.pdf GitHub

Abstract:
Discussions of AI in education focus predominantly on student-facing tools -- chatbots, tutors, and problem generators -- while the potential for the same infrastructure to support instructors remains largely unexplored. We describe Stan, a suite of tools for an undergraduate chemical engineering thermodynamics course built on a data pipeline that we develop and deploy in dual roles: serving students and supporting instructors from a shared foundation of lecture transcripts and a structured textbook index. On the student side, a retrieval-augmented generation (RAG) pipeline answers natural-language queries by extracting technical terms, matching them against the textbook index, and synthesizing grounded responses with specific chapter and page references. On the instructor side, the same transcript corpus is processed through structured analysis pipelines that produce per-lecture summaries, identify student questions and moments of confusion, and catalog the anecdotes and analogies used to motivate difficult material -- providing a searchable, semester-scale record of teaching that supports course reflection, reminders, and improvement. All components, including speech-to-text transcription, structured content extraction, and interactive query answering, run entirely on locally controlled hardware using open-weight models (Whisper large-v3, Llama~3.1 8B) with no dependence on cloud APIs, ensuring predictable costs, full data privacy, and reproducibility independent of third-party services. We describe the design, implementation, and practical failure modes encountered when deploying 7--8 billion parameter models for structured extraction over long lecture transcripts, including context truncation, bimodal output distributions, and schema drift, along with the mitigations that resolved them.

Paperid: 136, https://arxiv.org/pdf/2603.04597.pdf GitHub

Abstract:
Large language models (LLMs) typically receive diverse natural language (NL) feedback through interaction with the environment. However, current reinforcement learning (RL) algorithms rely solely on scalar rewards, leaving the rich information in NL feedback underutilized and leading to inefficient exploration. In this work, we propose GOLF, an RL framework that explicitly exploits group-level language feedback to guide targeted exploration through actionable refinements. GOLF aggregates two complementary feedback sources: (i) external critiques that pinpoint errors or propose targeted fixes, and (ii) intra-group attempts that supply alternative partial ideas and diverse failure patterns. These group-level feedbacks are aggregated to produce high-quality refinements, which are adaptively injected into training as off-policy scaffolds to provide targeted guidance in sparse-reward regions. Meanwhile, GOLF jointly optimizes generation and refinement within a unified RL loop, creating a virtuous cycle that continuously improves both capabilities. Experiments on both verifiable and non-verifiable benchmarks show that GOLF achieves superior performance and exploration efficiency, achieving 2.2$\times$ improvements in sample efficiency compared to RL methods trained solely on scalar rewards. Code is available at https://github.com/LuckyyySTA/GOLF.

Paperid: 137, https://arxiv.org/pdf/2603.04592.pdf GitHub

Abstract:
Standard Large Language Models (LLMs) are predominantly designed for static inference with pre-defined inputs, which limits their applicability in dynamic, real-time scenarios. To address this gap, the streaming LLM paradigm has emerged. However, existing definitions of streaming LLMs remain fragmented, conflating streaming generation, streaming inputs, and interactive streaming architectures, while a systematic taxonomy is still lacking. This paper provides a comprehensive overview and analysis of streaming LLMs. First, we establish a unified definition of streaming LLMs based on data flow and dynamic interaction to clarify existing ambiguities. Building on this definition, we propose a systematic taxonomy of current streaming LLMs and conduct an in-depth discussion on their underlying methodologies. Furthermore, we explore the applications of streaming LLMs in real-world scenarios and outline promising research directions to support ongoing advances in streaming intelligence. We maintain a continuously updated repository of relevant papers at https://github.com/EIT-NLP/Awesome-Streaming-LLMs.

Paperid: 138, https://arxiv.org/pdf/2603.04562.pdf GitHub

Abstract:
Local Climate Zones (LCZs) give a zoning map to study urban structures and land use and analyze the impact of urbanization on local climate. Multimodal remote sensing enables LCZ classification, for which data fusion is significant for improving accuracy owing to the data complexity. However, there is a gap in a comprehensive analysis of the fusion mechanisms used in their deep learning (DL) classifier architectures. This study analyzes different fusion strategies in the multi-class LCZ classification models for multimodal data and grouping strategies based on inherent data characteristics. The different models involving Convolutional Neural Networks (CNNs) include: (i) baseline hybrid fusion (FM1), (ii) with self- and cross-attention mechanisms (FM2), (iii) with the multi-scale Gaussian filtered images (FM3), and (iv) weighted decision-level fusion (FM4). Ablation experiments are conducted to study the pixel-, feature-, and decision-level fusion effects in the model performance. Grouping strategies include band grouping (BG) within the data modalities and label merging (LM) in the ground truth. Our analysis is exclusively done on the So2Sat LCZ42 dataset, which consists of Synthetic Aperture Radar (SAR) and Multispectral Imaging (MSI) image pairs. Our results show that FM1 consistently outperforms simple fusion methods. FM1 with BG and LM is found to be the most effective approach among all fusion strategies, giving an overall accuracy of 76.6\%. Importantly, our study highlights the effect of these strategies in improving prediction accuracy for the underrepresented classes. Our code and processed datasets are available at https://github.com/GVCL/LCZC-MultiModalHybridFusion

Paperid: 139, https://arxiv.org/pdf/2603.04553.pdf GitHub

Abstract:
We introduce Latent Particle World Model (LPWM), a self-supervised object-centric world model scaled to real-world multi-object datasets and applicable in decision-making. LPWM autonomously discovers keypoints, bounding boxes, and object masks directly from video data, enabling it to learn rich scene decompositions without supervision. Our architecture is trained end-to-end purely from videos and supports flexible conditioning on actions, language, and image goals. LPWM models stochastic particle dynamics via a novel latent action module and achieves state-of-the-art results on diverse real-world and synthetic datasets. Beyond stochastic video modeling, LPWM is readily applicable to decision-making, including goal-conditioned imitation learning, as we demonstrate in the paper. Code, data, pre-trained models and video rollouts are available: https://taldatech.github.io/lpwm-web

Paperid: 140, https://arxiv.org/pdf/2603.04532.pdf GitHub

Abstract:
Information retrieval (IR) benchmarks typically follow the Cranfield paradigm, relying on static and predefined corpora. However, temporal changes in technical corpora, such as API deprecations and code reorganizations, can render existing benchmarks stale. In our work, we investigate how temporal corpus drift affects FreshStack, a retrieval benchmark focused on technical domains. We examine two independent corpus snapshots of FreshStack from October 2024 and October 2025 to answer questions about LangChain. Our analysis shows that all but one query posed in 2024 remain fully supported by the 2025 corpus, as relevant documents "migrate" from LangChain to competitor repositories, such as LlamaIndex. Next, we compare the accuracy of retrieval models on both snapshots and observe only minor shifts in model rankings, with overall strong correlation of up to 0.978 Kendall $τ$ at Recall@50. These results suggest that retrieval benchmarks re-judged with evolving temporal corpora can remain reliable for retrieval evaluation. We publicly release all our artifacts at https://github.com/fresh-stack/driftbench.

Paperid: 141, https://arxiv.org/pdf/2603.04531.pdf GitHub

Abstract:
Tactile dexterous manipulation is essential to automating complex household tasks, yet learning effective control policies remains a challenge. While recent work has relied on imitation learning, obtaining high quality demonstrations for multi-fingered hands via robot teleoperation or kinesthetic teaching is prohibitive. Alternatively, with reinforcement we can learn skills in simulation, but fast and realistic simulation of tactile observations is challenging. To bridge this gap, we introduce PTLD: sim-to-real Privileged Tactile Latent Distillation, a novel approach to learning tactile manipulation skills without requiring tactile simulation. Instead of simulating tactile sensors or relying purely on proprioceptive policies to transfer zero-shot sim-to-real, our key idea is to leverage privileged sensors in the real world to collect real-world tactile policy data. This data is then used to distill a robust state estimator that operates on tactile input. We demonstrate from our experiments that PTLD can be used to improve proprioceptive manipulation policies trained in simulation significantly by incorporating tactile sensing. On the benchmark in-hand rotation task, PTLD achieves a 182% improvement over a proprioception only policy. We also show that PTLD enables learning the challenging task of tactile in-hand reorientation where we see a 57% improvement in the number of goals reached over using proprioception alone. Website: https://akashsharma02.github.io/ptld-website/.

Paperid: 142, https://arxiv.org/pdf/2603.04480.pdf GitHub

Abstract:
Machine learning-based antibody design is emerging as one of the most promising approaches to combat infectious diseases, due to significant advancements in the field of artificial intelligence and an exponential surge in experimental antibody data (in particular related to COVID-19). The ability of an antibody to bind to an antigens (called binding affinity) is one of the the most critical properties in designing neutralizing antibodies. In this study we introduce Ab-Affinity, a new large language model that can accurately predict the binding affinity of antibodies against a target peptide, e.g., the SARS-CoV-2 spike protein. Code and model are available at https://github.com/ucrbioinfo/AbAffinity.

Paperid: 143, https://arxiv.org/pdf/2603.04455.pdf GitHub

Abstract:
This paper investigates the integration of large language models (LLMs) as reasoning agents in repeated spectrum auctions within heterogeneous networks (HetNets). While auction-based mechanisms have been widely employed for efficient resource allocation, most prior works assume one-shot auctions, static bidder behavior, and idealized conditions. In contrast to traditional formulations where base station (BS) association and power allocation are centrally optimized, we propose a distributed auction-based framework in which each BS independently conducts its own multi-channel auction, and user equipments (UEs) strategically decide both their association and bid values. Within this setting, UEs operate under budget constraints and repeated interactions, transforming resource allocation into a long-term economic decision rather than a one-shot optimization problem. The proposed framework enables the evaluation of diverse bidding behaviors -from classical myopic and greedy policies to LLM-based agents capable of reasoning over historical outcomes, anticipating competition, and adapting their bidding strategy across episodes. Simulation results reveal that the LLM-empowered UE consistently achieves higher channel access frequency and improved budget efficiency compared to benchmarks. These findings highlight the potential of reasoning-enabled agents in future decentralized wireless networks markets and pave the way for lightweight, edge-deployable LLMs to support intelligent resource allocation in next-generation HetNets.

Paperid: 144, https://arxiv.org/pdf/2603.04454.pdf GitHub

Abstract:
How carefully and unambiguously a question is phrased has a profound impact on the quality of the response, for Language Models (LMs) as well as people. While model capabilities continue to advance, the interplay between grounding context and query formulation remains under-explored. This work investigates how the quality of background grounding information in a model's context window affects accuracy. We find that combining well-grounded dynamic context construction (i.e, RAG) with query rewriting reduces question ambiguity, resulting in significant accuracy gains. Given a user question with associated answer-free grounding context, rewriting the question to reduce ambiguity produces benchmark improvements without changing the answer itself, even compared to prepending that context before the question. Using \texttt{gpt-oss-20b} to rewrite a subset of Humanity's Last Exam using answer-free grounding context improves \texttt{gpt-5-mini} accuracy from 0.14 to 0.37. We demonstrate that this accuracy improvement cannot be fully recovered just through prompting at inference time; rather, distinct rewriting and answering phases are required. Code and data are available at https://github.com/mmajurski/lm-rewrite-uplift

Paperid: 145, https://arxiv.org/pdf/2603.04428.pdf GitHub

Abstract:
Multi-agent LLM systems on edge devices face a memory management problem: device RAM is too small to hold every agent's KV cache simultaneously. On Apple M4 Pro with 10.2 GB of cache budget, only 3 agents fit at 8K context in FP16. A 10-agent workflow must constantly evict and reload caches. Without persistence, every eviction forces a full re-prefill through the model -- 15.7 seconds per agent at 4K context. We address this by persisting each agent's KV cache to disk in 4-bit quantized format and reloading it directly into the attention layer, eliminating redundant O(n) prefill computation via direct cache restoration. The system comprises three components: a block pool providing per-agent isolated Q4 KV caches in safetensors format, a BatchQuantizedKVCache for concurrent inference over multiple agents' quantized caches, and cross-phase context injection that accumulates attention state across conversation phases without re-computation. Evaluated on three architectures (Gemma 3 12B, dense GQA, 48 layers; DeepSeek-Coder-V2-Lite 16B, MoE MLA, 27 layers; Llama 3.1 8B, dense GQA, 32 layers), cache restoration reduces time-to-first-token by up to 136x (Gemma: 22--136x at 4K--32K; DeepSeek: 11--76x at 4K--32K; Llama: 24--111x at 4K--16K; 3--10x at 1K). Q4 quantization fits 4x more agent contexts into fixed device memory than FP16. Perplexity measured with actual Q4 KV caches shows -0.7% for Gemma, +2.8% for Llama, and +3.0% for DeepSeek. Open-source at https://github.com/yshk-mxim/agent-memory

Paperid: 146, https://arxiv.org/pdf/2603.04419.pdf GitHub

Abstract:
We characterize the phenomenon of context-dependent affordance computation in vision-language models (VLMs). Through a large-scale computational study (n=3,213 scene-context pairs from COCO-2017) using Qwen-VL 30B and LLaVA-1.5-13B subject to systematic context priming across 7 agentic personas, we demonstrate massive affordance drift: mean Jaccard similarity between context conditions is 0.095 (95% CI: [0.093, 0.096], p < 0.0001), indicating that >90% of lexical scene description is context-dependent. Sentence-level cosine similarity confirms substantial drift at the semantic level (mean = 0.415, 58.5% context-dependent). Stochastic baseline experiments (2,384 inference runs across 4 temperatures and 5 seeds) confirm this drift reflects genuine context effects rather than generation noise: within-prime variance is substantially lower than cross-prime variance across all conditions. Tucker decomposition with bootstrap stability analysis (n=1,000 resamples) reveals stable orthogonal latent factors: a "Culinary Manifold" isolated to chef contexts and an "Access Axis" spanning child-mobility contrasts. These findings establish that VLMs compute affordances in a substantially context-dependent manner -- with the difference between lexical (90%) and semantic (58.5%) measures reflecting that surface vocabulary changes more than underlying meaning under context shifts -- and suggest a direction for robotics research: dynamic, query-dependent ontological projection (JIT Ontology) rather than static world modeling. We do not claim to establish processing order or architectural primacy; such claims require internal representational analysis beyond output behavior.

Paperid: 147, https://arxiv.org/pdf/2603.04415.pdf GitHub

Abstract:
While reasoning-enhanced Large Language Models (LLMs) have demonstrated remarkable advances in complex tasks such as mathematics and coding, their effectiveness across universal multimodal scenarios remains uncertain. The trend of releasing parallel "Instruct" and "Thinking" models by leading developers serves merely as a resource-intensive workaround, stemming from the lack of a criterion for determining when reasoning is truly beneficial. In this paper, we propose Dual Tuning, a framework designed to assess whether reasoning yields positive gains for target tasks under given base models and datasets. By jointly fine-tuning on paired Chain-of-Thought (CoT) and Direct-Answer (DA) data under controlled prompts, we systematically quantify and compare the gains of both training modes using the proposed metrics, and establish the "Thinking Boundary" to evaluate the suitability of reasoning training across diverse multimodal tasks, including spatial, mathematical, and multi-disciplinary domains. We further explore the impact of reinforcement training and thinking patterns on reasoning suitability, and validate whether the "Thinking Boundary" can guide data refinement. Our findings challenge the "reasoning-for-all" paradigm, providing practical guidance for identifying appropriate data and training strategies, and motivating the development of resource-efficient, adaptive auto-think systems.

Paperid: 148, https://arxiv.org/pdf/2603.04405.pdf GitHub

Abstract:
Foundation models are increasingly applied to computational pathology, yet their behavior under cross-cancer and cross-species transfer remains unspecified. This study investigated how fine-tuning CPath-CLIP affects cancer detection under same-cancer, cross-cancer, and cross-species conditions using whole-slide image patches from canine and human histopathology. Performance was measured using area under the receiver operating characteristic curve (AUC). Few-shot fine-tuning improved same-cancer (64.9% to 72.6% AUC) and cross-cancer performance (56.84% to 66.31% AUC). Cross-species evaluation revealed that while tissue matching enables meaningful transfer, performance remains below state-of-the-art benchmarks (H-optimus-0: 84.97% AUC), indicating that standard vision-language alignment is suboptimal for cross-species generalization. Embedding space analysis revealed extremely high cosine similarity (greater than 0.99) between tumor and normal prototypes. Grad-CAM shows prototype-based models remain domain-locked, while language-guided models attend to conserved tumor morphology. To address this, we introduce Semantic Anchoring, which uses language to provide a stable coordinate system for visual features. Ablation studies reveal that benefits stem from the text-alignment mechanism itself, regardless of text encoder complexity. Benchmarking against H-optimus-0 shows that CPath-CLIP's failure stems from intrinsic embedding collapse, which text alignment effectively circumvents. Additional gains were observed in same-cancer (8.52%) and cross-cancer classification (5.67%). We identified a previously uncharacterized failure mode: semantic collapse driven by species-dominated alignment rather than missing visual information. These results demonstrate that language acts as a control mechanism, enabling semantic re-interpretation without retraining.

Paperid: 149, https://arxiv.org/pdf/2603.04402.pdf GitHub

Abstract:
The rapid growth of Retrieval-Augmented Generation (RAG) has created a proliferation of toolkits, yet a fundamental gap remains between experimental prototypes and robust, production-ready systems. We present SearchGym, a modular infrastructure designed for cross-platform benchmarking and hybrid search orchestration. Unlike existing model-centric frameworks, SearchGym decouples data representation, embedding strategies, and retrieval logic into stateful abstractions: Dataset, VectorSet, and App. This separation enables a Compositional Config Algebra, allowing designers to synthesize entire systems from hierarchical configurations while ensuring perfect reproducibility. Moreover, we analyze the "Top-$k$ Cognizance" in hybrid retrieval pipelines, demonstrating that the optimal sequence of semantic ranking and structured filtering is highly dependent on filter strength. Evaluated on the LitSearch expert-annotated benchmark, SearchGym achieves a 70% Top-100 retrieval rate. SearchGym reveals a design tension between generalizability and optimizability, presenting the potential where engineering optimization may serve as a tool for uncovering the causal mechanisms inherent in information retrieval across heterogeneous domains. An open-source implementation of SearchGym is available at: https://github.com/JeromeTH/search-gym

Paperid: 150, https://arxiv.org/pdf/2603.04385.pdf GitHub

Abstract:
Feed-forward transformer models have driven rapid progress in 3D vision, but state-of-the-art methods such as VGGT and $π^3$ have a computational cost that scales quadratically with the number of input images, making them inefficient when applied to large image collections. Sequential-reconstruction approaches reduce this cost but sacrifice reconstruction quality. We introduce ZipMap, a stateful feed-forward model that achieves linear-time, bidirectional 3D reconstruction while matching or surpassing the accuracy of quadratic-time methods. ZipMap employs test-time training layers to zip an entire image collection into a compact hidden scene state in a single forward pass, enabling reconstruction of over 700 frames in under 10 seconds on a single H100 GPU, more than $20\times$ faster than state-of-the-art methods such as VGGT. Moreover, we demonstrate the benefits of having a stateful representation in real-time scene-state querying and its extension to sequential streaming reconstruction.

Paperid: 151, https://arxiv.org/pdf/2603.04384.pdf GitHub

Abstract:
Deep Research agents are rapidly emerging as primary consumers of modern retrieval systems. Unlike human users who issue and refine queries without documenting their intermediate thought processes, Deep Research agents generate explicit natural language reasoning before each search call, revealing rich intent and contextual information that existing retrievers entirely ignore. To exploit this overlooked signal, we introduce: (1) Reasoning-Aware Retrieval, a retrieval paradigm that jointly embeds the agent's reasoning trace alongside its query; and (2) DR-Synth, a data synthesis method that generates Deep Research retriever training data from standard QA datasets. We demonstrate that both components are independently effective, and their combination yields a trained embedding model, AgentIR-4B, with substantial gains. On the challenging BrowseComp-Plus benchmark, AgentIR-4B achieves 68\% accuracy with the open-weight agent Tongyi-DeepResearch, compared to 50\% with conventional embedding models twice its size, and 37\% with BM25. Code and data are available at: https://texttron.github.io/AgentIR/.

Paperid: 152, https://arxiv.org/pdf/2603.04366.pdf GitHub

Abstract:
Generative audio requires fine-grained controllable outputs, yet most existing methods require model retraining on specific controls or inference-time controls (\textit{e.g.}, guidance) that can also be computationally demanding. By examining the bottlenecks of existing guidance-based controls, in particular their high cost-per-step due to decoder backpropagation, we introduce a guidance-based approach through selective TFG and Latent-Control Heads (LatCHs), which enables controlling latent audio diffusion models with low computational overhead. LatCHs operate directly in latent space, avoiding the expensive decoder step, and requiring minimal training resources (7M parameters and $\approx$ 4 hours of training). Experiments with Stable Audio Open demonstrate effective control over intensity, pitch, and beats (and a combination of those) while maintaining generation quality. Our method balances precision and audio fidelity with far lower computational costs than standard end-to-end guidance. Demo examples can be found at https://zacharynovack.github.io/latch/latch.html.

Paperid: 153, https://arxiv.org/pdf/2603.04346.pdf GitHub

Abstract:
Large-scale Vision-Language Foundation Models (VLFMs), such as CLIP, now underpin a wide range of computer vision research and applications. VLFMs are often adapted to various domain-specific tasks. However, VLFM performance on novel, specialised, or underrepresented domains remains inconsistent. Evaluating VLFMs typically requires labelled test sets, which are often unavailable for niche domains of interest, particularly those from the Global South. We address this gap by proposing a highly data-efficient method to predict a VLFM's zero-shot accuracy on a target domain using only a single labelled image per class. Our approach uses a Large Language Model to generate plausible counterfactual descriptions of a given image. By measuring the VLFM's ability to distinguish the correct description from these hard negatives, we engineer features that capture the VLFM's discriminative power in its shared embedding space. A linear regressor trained on these similarity scores estimates the VLFM's zero-shot test accuracy across various visual domains with a Pearson-r correlation of 0.96. We demonstrate our method's performance across five diverse datasets, including standard benchmark datasets and underrepresented datasets from Africa. Our work provides a low-cost, reliable tool for probing VLFMs, enabling researchers and practitioners to make informed decisions about data annotation efforts before committing significant resources. The model training code, generated captions and counterfactuals are released here: https://github.com/chris-vorster/PreLabellingProbe.

Paperid: 154, https://arxiv.org/pdf/2603.04341.pdf GitHub

Abstract:
In many CLIP adaptation methods, a blending ratio hyperparameter controls the trade-off between general pretrained CLIP knowledge and the limited, dataset-specific supervision from the few-shot cases. Most few-shot CLIP adaptation techniques report results by ablation of the blending ratio on the test set or require additional validation sets to select the blending ratio per dataset, and thus are not strictly few-shot. We present a simple, validation-free method for learning the blending ratio in CLIP adaptation. Hold-One-Shot-Out (HOSO) presents a novel approach for CLIP-Adapter-style methods to compete in the newly established validation-free setting. CLIP-Adapter with HOSO (HOSO-Adapter) learns the blending ratio using a one-shot, hold-out set, while the adapter trains on the remaining few-shot support examples. Under the validation-free few-shot protocol, HOSO-Adapter outperforms the CLIP-Adapter baseline by more than 4 percentage points on average across 11 standard few-shot datasets. Interestingly, in the 8- and 16-shot settings, HOSO-Adapter outperforms CLIP-Adapter even with the optimal blending ratio selected on the test set. Ablation studies validate the use of a one-shot hold-out mechanism, decoupled training, and improvements over the naively learnt blending ratio baseline. Code is released here: https://github.com/chris-vorster/HOSO-Adapter

Paperid: 155, https://arxiv.org/pdf/2603.04323.pdf GitHub

Abstract:
Federated learning (FL) faces two structural tensions: gradient sharing enables data-reconstruction attacks, while non-IID client distributions degrade aggregation quality. We introduce PTOPOFL, a framework that addresses both challenges simultaneously by replacing gradient communication with topological descriptors derived from persistent homology (PH). Clients transmit only 48-dimensional PH feature vectors-compact shape summaries whose many-to-one structure makes inversion provably ill-posed-rather than model gradients. The server performs topology-guided personalised aggregation: clients are clustered by Wasserstein similarity between their PH diagrams, intra-cluster models are topology-weighted,and clusters are blended with a global consensus. We prove an information-contraction theorem showing that PH descriptors leak strictly less mutual information per sample than gradients under strongly convex loss functions, and we establish linear convergence of the Wasserstein-weighted aggregation scheme with an error floor strictly smaller than FedAvg. Evaluated against FedAvg, FedProx, SCAFFOLD, and pFedMe on a non-IID healthcare scenario (8 hospitals, 2 adversarial) and a pathological benchmark (10 clients), PTOPOFL achieves AUC 0.841 and 0.910 respectively-the highest in both settings-while reducing reconstruction risk by a factor of 4.5 relative to gradient sharing. Code is publicly available at https://github.com/MorillaLab/TopoFederatedL and data at https://doi.org/10.5281/zenodo.18827595.

Paperid: 156, https://arxiv.org/pdf/2603.04314.pdf GitHub

Abstract:
Animal re-identification (ReID) faces critical challenges due to viewpoint variations, particularly in Aerial-Ground (AG-ReID) settings where models must match individuals across drastic elevation changes. However, existing datasets lack the precise angular annotations required to systematically analyze these geometric variations. To address this, we introduce the Multi-view Oriented Observation (MOO) dataset, a large-scale synthetic AG-ReID dataset of $1,000$ cattle individuals captured from $128$ uniformly sampled viewpoints ($128,000$ annotated images). Using this controlled dataset, we quantify the influence of elevation and identify a critical elevation threshold, above which models generalize significantly better to unseen views. Finally, we validate the transferability to real-world applications in both zero-shot and supervised settings, demonstrating performance gains across four real-world cattle datasets and confirming that synthetic geometric priors effectively bridge the domain gap. Collectively, this dataset and analysis lay the foundation for future model development in cross-view animal ReID. MOO is publicly available at https://github.com/TurtleSmoke/MOO.

Paperid: 157, https://arxiv.org/pdf/2603.04308.pdf GitHub

Abstract:
Post-training quantization (PTQ) of transformers is known to suffer from severe accuracy degradation due to structured activation outliers, as originally analyzed by Bondarenko et al. (EMNLP 2021) in work associated with Qualcomm AI Research. This paper provides a reproducible empirical reproduction and systems-level extension of that phenomenon in BERT-base fine-tuned on QNLI. When global W8A8 quantization is applied, validation accuracy drops sharply from 89.66% (FP32) to 54.33%, a decrease of 35.33 points. Statistical analysis of FP32 activations shows strongly heavy-tailed behavior that intensifies with model depth: kurtosis reaches 271 in the final layers and approximately 55% of activation energy is concentrated in the top 1% of channels. We evaluate several mitigation strategies. Mixed precision PTQ restores accuracy close to the FP32 baseline (89.42%). Per-embedding-group (PEG) quantization shows strong sensitivity to grouping structure, improving accuracy from 66.12% with three groups to 86.18% with four groups. In contrast, percentile-based calibration, even at thresholds between 99.0 and 99.99, fails to recover accuracy (about 50.54%), indicating that large activation channels encode structured signal rather than rare noise. Deployment profiling on an RTX 3050 GPU shows minimal differences in latency and memory usage across methods (median latency about 58-59 ms; VRAM usage about 484-486 MB), highlighting the importance of hardware-aware evaluation. Overall, the results show that PTQ failure in transformers is primarily driven by structured channel dominance amplified through residual connections. Effective mitigation therefore requires channel-aware precision allocation rather than scalar clipping alone.

Paperid: 158, https://arxiv.org/pdf/2603.04293.pdf GitHub

Abstract:
The advancement of Machine learning (ML), Large Audio Language Models (LALMs), and autonomous AI agents in Music Information Retrieval (MIR) necessitates a shift from static tagging to rich, human-aligned representation learning. However, the scarcity of open-source infrastructure capable of capturing the subjective nuances of audio annotation remains a critical bottleneck. This paper introduces \textbf{LabelBuddy}, an open-source collaborative auto-tagging audio annotation tool designed to bridge the gap between human intent and machine understanding. Unlike static tools, it decouples the interface from inference via containerized backends, allowing users to plug in custom models for AI-assisted pre-annotation. We describe the system architecture, which supports multi-user consensus, containerized model isolation, and a roadmap for extending agents and LALMs. Code available at https://github.com/GiannisProkopiou/gsoc2022-Label-buddy.

Paperid: 159, https://arxiv.org/pdf/2603.04291.pdf GitHub

Abstract:
Generating high-quality 360° panoramic videos from perspective input is one of the crucial applications for virtual reality (VR), whereby high-resolution videos are especially important for immersive experience. Existing methods are constrained by computational limitations of vanilla diffusion models, only supporting $\leq$ 1K resolution native generation and relying on suboptimal post super-resolution to increase resolution. We introduce CubeComposer, a novel spatio-temporal autoregressive diffusion model that natively generates 4K-resolution 360° videos. By decomposing videos into cubemap representations with six faces, CubeComposer autoregressively synthesizes content in a well-planned spatio-temporal order, reducing memory demands while enabling high-resolution output. Specifically, to address challenges in multi-dimensional autoregression, we propose: (1) a spatio-temporal autoregressive strategy that orchestrates 360° video generation across cube faces and time windows for coherent synthesis; (2) a cube face context management mechanism, equipped with a sparse context attention design to improve efficiency; and (3) continuity-aware techniques, including cube-aware positional encoding, padding, and blending to eliminate boundary seams. Extensive experiments on benchmark datasets demonstrate that CubeComposer outperforms state-of-the-art methods in native resolution and visual quality, supporting practical VR application scenarios. Project page: https://lg-li.github.io/project/cubecomposer

Paperid: 160, https://arxiv.org/pdf/2603.04284.pdf GitHub

Abstract:
Autonomous robotic systems are increasingly deployed for mapping, monitoring, and inspection in complex and unstructured environments. However, most existing path planning approaches remain domain-specific (i.e., either on air, land, or sea), limiting their scalability and cross-platform applicability. This article presents OmniPlanner, a unified planning framework for autonomous exploration and inspection across aerial, ground, and underwater robots. The method integrates volumetric exploration and viewpoint-based inspection, alongside target reach behaviors within a single modular architecture, complemented by a platform abstraction layer that captures morphology-specific sensing, traversability and motion constraints. This enables the same planning strategy to generalize across distinct mobility domains with minimal retuning. The framework is validated through extensive simulation studies and field deployments in underground mines, industrial facilities, forests, submarine bunkers, and structured outdoor environments. Across these diverse scenarios, OmniPlanner demonstrates robust performance, consistent cross-domain generalization, and improved exploration and inspection efficiency compared to representative state-of-the-art baselines.

Paperid: 161, https://arxiv.org/pdf/2603.04254.pdf GitHub GitHub

Abstract:
Understanding a 3D scene immediately with its exploration is essential for embodied tasks, where an agent must construct and comprehend the 3D scene in an online and nearly real-time manner. In this study, we propose EmbodiedSplat, an online feed-forward 3DGS for open-vocabulary scene understanding that enables simultaneous online 3D reconstruction and 3D semantic understanding from the streaming images. Unlike existing open-vocabulary 3DGS methods which are typically restricted to either offline or per-scene optimization setting, our objectives are two-fold: 1) Reconstructs the semantic-embedded 3DGS of the entire scene from over 300 streaming images in an online manner. 2) Highly generalizable to novel scenes with feed-forward design and supports nearly real-time 3D semantic reconstruction when combined with real-time 2D models. To achieve these objectives, we propose an Online Sparse Coefficients Field with a CLIP Global Codebook where it binds the 2D CLIP embeddings to each 3D Gaussian while minimizing memory consumption and preserving the full semantic generalizability of CLIP. Furthermore, we generate 3D geometric-aware CLIP features by aggregating the partial point cloud of 3DGS through 3D U-Net to compensate the 3D geometric prior to 2D-oriented language embeddings. Extensive experiments on diverse indoor datasets, including ScanNet, ScanNet++, and Replica, demonstrate both the effectiveness and efficiency of our method. Check out our project page in https://0nandon.github.io/EmbodiedSplat/.

Paperid: 162, https://arxiv.org/pdf/2603.04240.pdf GitHub

Abstract:
Pathology Foundation Models (FMs) have shown strong performance across a wide range of pathology image representation and diagnostic tasks. However, FMs do not exhibit the expected performance advantage over traditional specialized models in Nuclei Detection and Classification (NDC). In this work, we reveal that jointly optimizing nuclei detection and classification leads to severe representation degradation in FMs. Moreover, we identify that the substantial intrinsic disparity in task difficulty between nuclei detection and nuclei classification renders joint NDC optimization unnecessarily computationally burdensome for the detection stage. To address these challenges, we propose DeNuC, a simple yet effective method designed to break through existing bottlenecks by Decoupling Nuclei detection and Classification. DeNuC employs a lightweight model for accurate nuclei localization, subsequently leveraging a pathology FM to encode input images and query nucleus-specific features based on the detected coordinates for classification. Extensive experiments on three widely used benchmarks demonstrate that DeNuC effectively unlocks the representational potential of FMs for NDC and significantly outperforms state-of-the-art methods. Notably, DeNuC improves F1 scores by 4.2% and 3.6% (or higher) on the BRCAM2C and PUMA datasets, respectively, while using only 16% (or fewer) trainable parameters compared to other methods. Code is available at https://github.com/ZijiangY1116/DeNuC.

Paperid: 163, https://arxiv.org/pdf/2603.04239.pdf GitHub

Abstract:
Recent breakthroughs in Diffusion Transformers (DiTs) have revolutionized the field of visual synthesis due to their superior scalability. To facilitate DiTs' capability of capturing meaningful internal representations, recent works such as REPA incorporate external pretrained encoders for representation alignment. However, the underlying mechanisms governing representation learning within DiTs are not well understood. To this end, we first systematically investigate the representation dynamics of DiTs. Through analyzing the evolution and influence of internal representations under various settings, we reveal that representation diversity across blocks is a crucial factor for effective learning. Based on this key insight, we propose DiverseDiT, a novel framework that explicitly promotes representation diversity. DiverseDiT incorporates long residual connections to diversify input representations across blocks and a representation diversity loss to encourage blocks to learn distinct features. Extensive experiments on ImageNet 256x256 and 512x512 demonstrate that our DiverseDiT yields consistent performance gains and convergence acceleration when applied to different backbones with various sizes, even when tested on the challenging one-step generation setting. Furthermore, we show that DiverseDiT is complementary to existing representation learning techniques, leading to further performance gains. Our work provides valuable insights into the representation learning dynamics of DiTs and offers a practical approach for enhancing their performance.

Paperid: 164, https://arxiv.org/pdf/2603.04212.pdf GitHub

Abstract:
The rapid adoption of Large Language Models (LLMs) has transformed modern software development by enabling automated code generation at scale. While these systems improve productivity, they introduce new challenges for software governance, accountability, and compliance. Existing research primarily focuses on distinguishing machine-generated code from human-written code; however, many practical scenarios--such as vulnerability triage, incident investigation, and licensing audits--require identifying which LLM produced a given code snippet. In this paper, we study the problem of model-level code attribution, which aims to determine the source LLM responsible for generated code. Although attribution is challenging, differences in training data, architectures, alignment strategies, and decoding mechanisms introduce model-dependent stylistic and structural variations that serve as generative fingerprints. Leveraging this observation, we propose the Disentangled Code Attribution Network (DCAN), which separates Source-Agnostic semantic information from Source-Specific stylistic representations. Through a contrastive learning objective, DCAN isolates discriminative model-dependent signals while preserving task semantics, enabling multi-class attribution across models and programming languages. To support systematic evaluation, we construct the first large-scale benchmark dataset comprising code generated by four widely used LLMs (DeepSeek, Claude, Qwen, and ChatGPT) across four programming languages (Python, Java, C, and Go). Experimental results demonstrate that DCAN achieves reliable attribution performance across diverse settings, highlighting the feasibility of model-level provenance analysis in software engineering contexts. The dataset and implementation are publicly available at https://github.com/mtt500/DCAN.

Paperid: 165, https://arxiv.org/pdf/2603.04191.pdf GitHub

Abstract:
Large Language Models (LLMs) are increasingly serving as personal assistants, where users share complex and diverse preferences over extended interactions. However, assessing how well LLMs can follow these preferences in realistic, long-term situations remains underexplored. This work proposes RealPref, a benchmark for evaluating realistic preference-following in personalized user-LLM interactions. RealPref features 100 user profiles, 1300 personalized preferences, four types of preference expression (ranging from explicit to implicit), and long-horizon interaction histories. It includes three types of test questions (multiple-choice, true-or-false, and open-ended), with detailed rubrics for LLM-as-a-judge evaluation. Results indicate that LLM performance significantly drops as context length grows and preference expression becomes more implicit, and that generalizing user preference understanding to unseen scenarios poses further challenges. RealPref and these findings provide a foundation for future research to develop user-aware LLM assistants that better adapt to individual needs. The code is available at https://github.com/GG14127/RealPref.

Paperid: 166, https://arxiv.org/pdf/2603.04179.pdf GitHub

Abstract:
We present NOVA3R, an effective approach for non-pixel-aligned 3D reconstruction from a set of unposed images in a feed-forward manner. Unlike pixel-aligned methods that tie geometry to per-ray predictions, our formulation learns a global, view-agnostic scene representation that decouples reconstruction from pixel alignment. This addresses two key limitations in pixel-aligned 3D: (1) it recovers both visible and invisible points with a complete scene representation, and (2) it produces physically plausible geometry with fewer duplicated structures in overlapping regions. To achieve this, we introduce a scene-token mechanism that aggregates information across unposed images and a diffusion-based 3D decoder that reconstructs complete, non-pixel-aligned point clouds. Extensive experiments on both scene-level and object-level datasets demonstrate that NOVA3R outperforms state-of-the-art methods in terms of reconstruction accuracy and completeness.

Paperid: 167, https://arxiv.org/pdf/2603.04165.pdf GitHub GitHub

Abstract:
Large-scale 2D foundation models exhibit strong transferable representations, yet extending them to 3D volumetric data typically requires retraining, adapters, or architectural redesign. We introduce PlaneCycle, a training-free, adapter-free operator for architecture-agnostic 2D-to-3D lifting of foundation models. PlaneCycle reuses the original pretrained 2D backbone by cyclically distributing spatial aggregation across orthogonal HW, DW, and DH planes throughout network depth, enabling progressive 3D fusion while preserving pretrained inductive biases. The method introduces no additional parameters and is applicable to arbitrary 2D networks. Using pretrained DINOv3 models, we evaluate PlaneCycle on six 3D classification and three 3D segmentation benchmarks. Without any training, the lifted models exhibit intrinsic 3D fusion capability and, under linear probing, outperform slice-wise 2D baselines and strong 3D counterparts, approaching the performance of fully trained models. With full fine-tuning, PlaneCycle matches standard 3D architectures, highlighting its potential as a seamless and practical 2D-to-3D lifting operator. These results demonstrate that 3D capability can be unlocked from pretrained 2D foundation models without structural modification or retraining. Code is available at https://github.com/HINTLab/PlaneCycle.

Paperid: 168, https://arxiv.org/pdf/2603.04158.pdf GitHub

Abstract:
Garment manipulation has attracted increasing attention due to its critical role in home-assistant robotics. However, the majority of existing garment manipulation works assume an initial state consisting of only one garment, while piled garments are far more common in real-world settings. To bridge this gap, we propose a novel garment retrieval pipeline that can not only follow language instruction to execute safe and clean retrieval but also guarantee exactly one garment is retrieved per attempt, establishing a robust foundation for the execution of downstream tasks (e.g., folding, hanging, wearing). Our pipeline seamlessly integrates vision-language reasoning with visual affordance perception, fully leveraging the high-level reasoning and planning capabilities of VLMs alongside the generalization power of visual affordance for low-level actions. To enhance the VLM's comprehensive awareness of each garment's state within a garment pile, we employ visual segmentation model (SAM2) to execute object segmentation on the garment pile for aiding VLM-based reasoning with sufficient visual cues. A mask fine-tuning mechanism is further integrated to address scenarios where the initial segmentation results are suboptimal. In addition, a dual-arm cooperation framework is deployed to address cases involving large or long garments, as well as excessive garment sagging caused by incorrect grasping point determination, both of which are strenuous for a single arm to handle. The effectiveness of our pipeline are consistently demonstrated across diverse tasks and varying scenarios in both real-world and simulation environments. Project page: https://garmentpile2.github.io/.

Paperid: 169, https://arxiv.org/pdf/2603.04145.pdf GitHub

Abstract:
We present VietNormalizer1, an open-source, zero-dependency Python library for Vietnamese text normalization targeting Text-to-Speech (TTS) and Natural Language Processing (NLP) applications. Vietnamese text normalization is a critical yet underserved preprocessing step: real-world Vietnamese text is densely populated with non-standard words (NSWs), including numbers, dates, times, currency amounts, percentages, acronyms, and foreign-language terms, all of which must be converted to fully pronounceable Vietnamese words before TTS synthesis or downstream language processing. Existing Vietnamese normalization tools either require heavy neural dependencies while covering only a narrow subset of NSW classes, or are embedded within larger NLP toolkits without standalone installability. VietNormalizer addresses these gaps through a unified, rule-based pipeline that: (1) converts arbitrary integers, decimals, and large numbers to Vietnamese words; (2) normalizes dates and times to their spoken Vietnamese forms; (3) handles VND and USD currency amounts; (4) expands percentages; (5) resolves acronyms via a customizable CSV dictionary; (6) transliterates non-Vietnamese loanwords and foreign terms to Vietnamese phonetic approximations; and (7) performs Unicode normalization and emoji/special-character removal. All regular expression patterns are pre-compiled at initialization, enabling high-throughput batch processing with minimal memory overhead and no GPU or external API dependency. The library is installable via pip install vietnormalizer, available on PyPI and GitHub at https://github.com/nghimestudio/vietnormalizer, and released under the MIT license. We discuss the design decisions, limitations of existing approaches, and the generalizability of the rule-based normalization paradigm to other low-resource tonal and agglutinative languages.

Paperid: 170, https://arxiv.org/pdf/2603.04125.pdf GitHub

Abstract:
Few-Shot Action Recognition (FS-AR) has shown promising results but is often limited by a closed-set assumption that fails in real-world open-set scenarios. While Few-Shot Open-Set (FSOS) recognition is well-established for images, its extension to spatio-temporal video data remains underexplored. To address this, we propose an architectural extension based on a Feature-Residual Discriminator (FR-Disc), adapting previous work on skeletal data to the more complex video domain. Extensive experiments on five datasets demonstrate that while common open-set techniques provide only marginal gains, our FR-Disc significantly enhances unknown rejection capabilities without compromising closed-set accuracy, setting a new state-of-the-art for FSOS-AR. The project website, code, and benchmark are available at: https://hsp-iit.github.io/fsosar/.

Paperid: 171, https://arxiv.org/pdf/2603.04114.pdf GitHub

Abstract:
Multi-modal remote sensing imagery provides complementary observations of the same geographic scene, yet such observations are frequently incomplete in practice. Existing cross-modal translation methods treat each modality pair as an independent task, resulting in quadratic complexity and limited generalization to unseen modality combinations. We formulate Any-to-Any translation as inference over a shared latent representation of the scene, where different modalities correspond to partial observations of the same underlying semantics. Based on this formulation, we propose Any2Any, a unified latent diffusion framework that projects heterogeneous inputs into a geometrically aligned latent space. Such structure performs anchored latent regression with a shared backbone, decoupling modality-specific representation learning from semantic mapping. Moreover, lightweight target-specific residual adapters are used to correct systematic latent mismatches without increasing inference complexity. To support learning under sparse but connected supervision, we introduce RST-1M, the first million-scale remote sensing dataset with paired observations across five sensing modalities, providing supervision anchors for any-to-any translation. Experiments across 14 translation tasks show that Any2Any consistently outperforms pairwise translation methods and exhibits strong zero-shot generalization to unseen modality pairs. Code and models will be available at https://github.com/MiliLab/Any2Any.

Paperid: 172, https://arxiv.org/pdf/2603.04099.pdf GitHub GitHub

Abstract:
Multi-Layer Perceptron (MLP) models are the foundation of contemporary point cloud processing. However, their complex network architectures obscure the source of their strength and limit the application of these models. In this article, we develop a two-stage abstraction and refinement (ABS-REF) view for modular feature extraction in point cloud processing. This view elucidates that whereas the early models focused on ABS stages, the more recent techniques devise sophisticated REF stages to attain performance advantages. Then, we propose a High-dimensional Positional Encoding (HPE) module to explicitly utilize intrinsic positional information, extending the ``positional encoding'' concept from Transformer literature. HPE can be readily deployed in MLP-based architectures and is compatible with transformer-based methods. Within our ABS-REF view, we rethink local aggregation in MLP-based methods and propose replacing time-consuming local MLP operations, which are used to capture local relationships among neighbors. Instead, we use non-local MLPs for efficient non-local information updates, combined with the proposed HPE for effective local information representation. We leverage our modules to develop HPENets, a suite of MLP networks that follow the ABS-REF paradigm, incorporating a scalable HPE-based REF stage. Extensive experiments on seven public datasets across four different tasks show that HPENets deliver a strong balance between efficiency and effectiveness. Notably, HPENet surpasses PointNeXt, a strong MLP-based counterpart, by 1.1% mAcc, 4.0% mIoU, 1.8% mIoU, and 0.2% Cls. mIoU, with only 50.0%, 21.5%, 23.1%, 44.4% of FLOPs on ScanObjectNN, S3DIS, ScanNet, and ShapeNetPart, respectively. Source code is available at https://github.com/zouyanmei/HPENet_v2.git.

Paperid: 173, https://arxiv.org/pdf/2603.04091.pdf GitHub

Abstract:
Modeling plant growth dynamics plays a central role in modern agricultural research. However, learning robust predictors from multi-view plant imagery remains challenging due to strong viewpoint redundancy and viewpoint-dependent appearance changes. We propose a level-aware vision language framework that jointly predicts plant age and leaf count using a single multi-task model built on CLIP embeddings. Our method aggregates rotational views into angle-invariant representations and conditions visual features on lightweight text priors encoding viewpoint level for stable prediction under incomplete or unordered inputs. On the GroMo25 benchmark, our approach reduces mean age MAE from 7.74 to 3.91 and mean leaf-count MAE from 5.52 to 3.08 compared to the GroMo baseline, corresponding to improvements of 49.5% and 44.2%, respectively. The unified formulation simplifies the pipeline by replacing the conventional dual-model setup while improving robustness to missing views. The models and code is available at: https://github.com/SimonWarmers/CLIP-MVP

Paperid: 174, https://arxiv.org/pdf/2603.04058.pdf GitHub

Abstract:
Glioblastoma exhibits diverse, infiltrative, and patient-specific growth patterns that are only partially visible on routine MRI, making it difficult to reliably assess true tumor extent and personalize treatment planning and follow-up. We present a biophysically-conditioned generative framework that synthesizes biologically realistic 3D brain MRI volumes from estimated, spatially continuous tumor-concentration fields. Our approach combines a generative model with tumor-infiltration maps that can be propagated through time using a biophysical growth model, enabling fine-grained control over tumor shape and growth while preserving patient anatomy. This enables us to synthesize consistent tumor growth trajectories directly in the space of real patients, providing interpretable, controllable estimation of tumor infiltration and progression beyond what is explicitly observed in imaging. We evaluate the framework on longitudinal glioblastoma cases and demonstrate that it can generate temporally coherent sequences with realistic changes in tumor appearance and surrounding tissue response. These results suggest that integrating mechanistic tumor growth priors with modern generative modeling can provide a practical tool for patient-specific progression visualization and for generating controlled synthetic data to support downstream neuro-oncology workflows. In longitudinal extrapolation, we achieve a consistent 75% Dice overlap with the biophysical model while maintaining a constant PSNR of 25 in the surrounding tissue. Our code is available at: https://github.com/valentin-biller/lgm.git

Paperid: 175, https://arxiv.org/pdf/2603.04050.pdf GitHub

Abstract:
In this work, we propose HE-VPR, a visual place recognition (VPR) framework that incorporates height estimation. Our system decouples height inference from place recognition, allowing both modules to share a frozen DINOv2 backbone. Two lightweight bypass adapter branches are integrated into our system. The first estimates the height partition of the query image via retrieval from a compact height database, and the second performs VPR within the corresponding height-specific sub-database. The adaptation design reduces training cost and significantly decreases the search space of the database. We also adopt a center-weighted masking strategy to further enhance the robustness against scale differences. Experiments on two self-collected challenging multi-altitude datasets demonstrate that HE-VPR achieves up to 6.1\% Recall@1 improvement over state-of-the-art ViT-based baselines and reduces memory usage by up to 90\%. These results indicate that HE-VPR offers a scalable and efficient solution for height-aware aerial VPR, enabling practical deployment in GNSS-denied environments. All the code and datasets for this work have been released on https://github.com/hmf21/HE-VPR.

Paperid: 176, https://arxiv.org/pdf/2603.04035.pdf GitHub

Abstract:
mlx-vis is a Python library that implements six dimensionality reduction methods and a k-nearest neighbor graph algorithm entirely in MLX, Apple's array framework for Apple Silicon. The library provides UMAP, t-SNE, PaCMAP, TriMap, DREAMS, CNE, and NNDescent, all executing on Metal GPU through a unified fit_transform interface. Beyond embedding computation, mlx-vis includes a GPU-accelerated circle-splatting renderer that produces scatter plots and smooth animations without matplotlib, composing frames via scatter-add alpha blending on GPU and piping them to hardware H.264 encoding. On Fashion-MNIST with 70,000 points, all methods complete embedding in 2.1-3.8 seconds and render 800-frame animations in 1.4 seconds on an M3 Ultra, with the full pipeline from raw data to rendered video finishing in 3.6-5.2 seconds. The library depends only on MLX and NumPy, is released under the Apache 2.0 license, and is available at https://github.com/hanxiao/mlx-vis.

Paperid: 177, https://arxiv.org/pdf/2603.04002.pdf GitHub

Abstract:
Reasoning segmentation increasingly employs reinforcement learning to generate explanatory reasoning chains that guide Multimodal Large Language Models. While these geometric rewards are primarily confined to guiding the final localization, they are incapable of discriminating whether the reasoning process remains anchored on the referred region or strays into irrelevant context. Lacking this discriminative guidance, the model's reasoning often devolves into unfocused and verbose chains that ultimately fail to disambiguate and perceive the target in complex scenes. This suggests a need to complement the RL objective with Discriminative Perception, an ability to actively distinguish a target from its context. To realize this, we propose DPAD to compel the model to generate a descriptive caption of the referred object, which is then used to explicitly discriminate by contrasting the caption's semantic relevance to the referred object against the wider context. By optimizing for this discriminative capability, the model is forced to focus on the unique attributes of the target, leading to a more converged and efficient reasoning chain. The descriptive caption also serves as an interpretability rationale that aligns with the segmentation. Experiments on the benchmarks confirm the validity of our approach, delivering substantial performance gains, with the cIoU on ReasonSeg increasing by 3.09% and the reasoning chain length decreasing by approximately 42%. Code is available at https://github.com/mrazhou/DPAD

Paperid: 178, https://arxiv.org/pdf/2603.03985.pdf GitHub

Abstract:
The rapid advancement of multimodal large language models has demonstrated impressive capabilities, yet nearly all operate in an offline paradigm, hindering real-time interactivity. Addressing this gap, we introduce the Real-tIme Video intERaction Bench (RIVER Bench), designed for evaluating online video comprehension. RIVER Bench introduces a novel framework comprising Retrospective Memory, Live-Perception, and Proactive Anticipation tasks, closely mimicking interactive dialogues rather than responding to entire videos at once. We conducted detailed annotations using videos from diverse sources and varying lengths, and precisely defined the real-time interactive format. Evaluations across various model categories reveal that while offline models perform well in single question-answering tasks, they struggle with real-time processing. Addressing the limitations of existing models in online video interaction, especially their deficiencies in long-term memory and future perception, we proposed a general improvement method that enables models to interact with users more flexibly in real time. We believe this work will significantly advance the development of real-time interactive video understanding models and inspire future research in this emerging field. Datasets and code are publicly available at https://github.com/OpenGVLab/RIVER.

Paperid: 179, https://arxiv.org/pdf/2603.03968.pdf GitHub

Abstract:
Nominal techniques provide a mathematically principled approach to dealing with names and variable binding in programming languages. This paper explores an attempt to make nominal techniques accessible as an Agda library. We aim for a technical victory of implementing nominal ideas; we further require a moral victory that the overhead be acceptable for practical systems. The results of this paper have been mechanised and are publicly accessible at https://omelkonian.github.io/nominal-agda/.

Paperid: 180, https://arxiv.org/pdf/2603.03967.pdf GitHub

Abstract:
Despite significant progress has been made in image deraining, we note that most existing methods are often developed for only specific types of rain degradation and fail to generalize across diverse real-world rainy scenes. How to effectively model different rain degradations within a universal framework is important for real-world image deraining. In this paper, we propose UniRain, an effective unified image deraining framework capable of restoring images degraded by rain streak and raindrop under both daytime and nighttime conditions. To better enhance unified model generalization, we construct an intelligent retrieval augmented generation (RAG)-based dataset distillation pipeline that selects high-quality training samples from all public deraining datasets for better mixed training. Furthermore, we incorporate a simple yet effective multi-objective reweighted optimization strategy into the asymmetric mixture-of-experts (MoE) architecture to facilitate consistent performance and improve robustness across diverse scenes. Extensive experiments show that our framework performs favorably against the state-of-the-art models on our proposed benchmarks and multiple public datasets.

Paperid: 181, https://arxiv.org/pdf/2603.03953.pdf GitHub

Abstract:
Safe visual navigation is critical for indoor mobile robots operating in cluttered environments. Existing benchmarks, however, often neglect collisions or are designed for outdoor scenarios, making them unsuitable for indoor visual navigation. To address this limitation, we introduce the reactive visual navigation benchmark (RVN-Bench), a collision-aware benchmark for indoor mobile robots. In RVN-Bench, an agent must reach sequential goal positions in previously unseen environments using only visual observations and no prior map, while avoiding collisions. Built on the Habitat 2.0 simulator and leveraging high-fidelity HM3D scenes, RVN-Bench provides large-scale, diverse indoor environments, defines a collision-aware navigation task and evaluation metrics, and offers tools for standardized training and benchmarking. RVN-Bench supports both online and offline learning by offering an environment for online reinforcement learning, a trajectory image dataset generator, and tools for producing negative trajectory image datasets that capture collision events. Experiments show that policies trained on RVN-Bench generalize effectively to unseen environments, demonstrating its value as a standardized benchmark for safe and robust visual navigation. Code and additional materials are available at: https://rvn-bench.github.io/.

Paperid: 182, https://arxiv.org/pdf/2603.03944.pdf GitHub

Abstract:
Spatial reasoning, the ability to understand spatial relations, causality, and dynamic evolution, is central to human intelligence and essential for real-world applications such as autonomous driving and robotics. Existing studies, however, primarily assess models on visible spatio-temporal understanding, overlooking their ability to infer unseen past or future spatial states. In this work, we introduce Spatial Causal Prediction (SCP), a new task paradigm that challenges models to reason beyond observation and predict spatial causal outcomes. We further construct SCP-Bench, a benchmark comprising 2,500 QA pairs across 1,181 videos spanning diverse viewpoints, scenes, and causal directions, to support systematic evaluation. Through comprehensive experiments on {23} state-of-the-art models, we reveal substantial gaps between human and model performance, limited temporal extrapolation, and weak causal grounding. We further analyze key factors influencing performance and propose perception-enhancement and reasoning-guided strategies toward advancing spatial causal intelligence. The project page is https://guangstrip.github.io/SCP-Bench.

Paperid: 183, https://arxiv.org/pdf/2603.03942.pdf GitHub

Abstract:
Robots operating in shared human environments must not only navigate, interact, and detect their surroundings, they must also interpret and respond to dynamic, and often unpredictable, human behaviours. Although recent advances have shown promise in enhancing robotic perception and instruction-following using Vision-Language Models (VLMs), they remain limited in addressing the complexities of multimodal human-robot interactions (HRI). Motivated by this challenge, we introduce a lightweight language-to-vision feedback module that closes the loop between an LLM and the vision encoder in VLMs. The module projects image-token hidden states through a gated Multi-Layer Perceptron (MLP) back into the encoder input, prompting a second pass that reinterprets the scene under text context. We evaluate this approach on three robotics-centred tasks: navigation in a simulated environment (Habitat), sequential scene description (Mementos-Robotics), and human-intention recognition (our HRI dataset). Results show that our method improves Qwen 2.5 (7B) by $3.3\%$ (less distance), $+0.057$ description score, and $+2.93\%$ accuracy, with less than $3\%$ extra parameters; Gemma 3 (4B) and LLaVA OV 1.5 (4B) show mixed navigation results but gains $+0.111,+0.055$ and $+10.81\%,+4.79\%$ on the latter two tasks. Code is available at https://github.com/alessioGalatolo/VLM-Reasoning-for-Robotics

Paperid: 184, https://arxiv.org/pdf/2603.03939.pdf GitHub

Abstract:
Multimodal industrial anomaly detection benefits from integrating RGB appearance with 3D surface geometry, yet existing \emph{unsupervised} approaches commonly rely on memory banks, teacher-student architectures, or fragile fusion schemes, limiting robustness under noisy depth, weak texture, or missing modalities. This paper introduces \textbf{CMDR-IAD}, a lightweight and modality-flexible unsupervised framework for reliable anomaly detection in 2D+3D multimodal as well as single-modality (2D-only or 3D-only) settings. \textbf{CMDR-IAD} combines bidirectional 2D$\leftrightarrow$3D cross-modal mapping to model appearance-geometry consistency with dual-branch reconstruction that independently captures normal texture and geometric structure. A two-part fusion strategy integrates these cues: a reliability-gated mapping anomaly highlights spatially consistent texture-geometry discrepancies, while a confidence-weighted reconstruction anomaly adaptively balances appearance and geometric deviations, yielding stable and precise anomaly localization even in depth-sparse or low-texture regions. On the MVTec 3D-AD benchmark, CMDR-IAD achieves state-of-the-art performance while operating without memory banks, reaching 97.3\% image-level AUROC (I-AUROC), 99.6\% pixel-level AUROC (P-AUROC), and 97.6\% AUPRO. On a real-world polyurethane cutting dataset, the 3D-only variant attains 92.6\% I-AUROC and 92.5\% P-AUROC, demonstrating strong effectiveness under practical industrial conditions. These results highlight the framework's robustness, modality flexibility, and the effectiveness of the proposed fusion strategies for industrial visual inspection. Our source code is available at https://github.com/ECGAI-Research/CMDR-IAD/

Paperid: 185, https://arxiv.org/pdf/2603.03935.pdf GitHub

Abstract:
Open-set semantic mapping enables language-driven robotic perception, but current instance-centric approaches are bottlenecked by context-depriving and computationally expensive crop-based feature extraction. To overcome this fundamental limitation, we introduce DISC (Dense Integrated Semantic Context), featuring a novel single-pass, distance-weighted extraction mechanism. By deriving high-fidelity CLIP embeddings directly from the vision transformer's intermediate layers, our approach eliminates the latency and domain-shift artifacts of traditional image cropping, yielding pure, mask-aligned semantic representations. To fully leverage these features in large-scale continuous mapping, DISC is built upon a fully GPU-accelerated architecture that replaces periodic offline processing with precise, on-the-fly voxel-level instance refinement. We evaluate our approach on standard benchmarks (Replica, ScanNet) and a newly generated large-scale-mapping dataset based on Habitat-Matterport 3D (HM3DSEM) to assess scalability across complex scenes in multi-story buildings. Extensive evaluations demonstrate that DISC significantly surpasses current state-of-the-art zero-shot methods in both semantic accuracy and query retrieval, providing a robust, real-time capable framework for robotic deployment. The full source code, data generation and evaluation pipelines will be made available at https://github.com/DFKI-NI/DISC.

Paperid: 186, https://arxiv.org/pdf/2603.03903.pdf GitHub GitHub

Abstract:
Building reliable classifiers is a fundamental challenge for deploying machine learning in real-world applications. A reliable system should not only detect out-of-distribution (OOD) inputs but also anticipate in-distribution (ID) errors by assigning low confidence to potentially misclassified samples. Yet, most prior work treats OOD detection and failure prediction as separated problems, overlooking their closed connection. We argue that reliability requires evaluating them jointly. To this end, we propose a unified evaluation framework that integrates OOD detection and failure prediction, quantified by our new metrics DS-F1 and DS-AURC, where DS denotes double scoring functions. Experiments on the OpenOOD benchmark show that double scoring functions yield classifiers that are substantially more reliable than traditional single scoring approaches. Our analysis further reveals that OOD-based approaches provide notable gains under simple or far-OOD shifts, but only marginal benefits under more challenging near-OOD conditions. Beyond evaluation, we extend the reliable classifier SURE and introduce SURE+, a new approach that significantly improves reliability across diverse scenarios. Together, our framework, metrics, and method establish a new benchmark for trustworthy classification and offer practical guidance for deploying robust models in real-world settings. The source code is publicly available at https://github.com/Intellindust-AI-Lab/SUREPlus.

Paperid: 187, https://arxiv.org/pdf/2603.03884.pdf GitHub

Abstract:
Topic localization aims to identify spans of text that express a given topic defined by a name and description. To study this task, we introduce a human-annotated benchmark based on Czech historical documents, containing human-defined topics together with manually annotated spans and supporting evaluation at both document and word levels. Evaluation is performed relative to human agreement rather than a single reference annotation. We evaluate a diverse range of large language models alongside BERT-based models fine-tuned on a distilled development dataset. Results reveal substantial variability among LLMs, with performance ranging from near-human topic detection to pronounced failures in span localization. While the strongest models approach human agreement, the distilled token embedding models remain competitive despite their smaller scale. The dataset and evaluation framework are publicly available at: https://github.com/dcgm/czechtopic.

Paperid: 188, https://arxiv.org/pdf/2603.03880.pdf GitHub

Abstract:
Software-hardware co-design is essential for optimizing in-memory computing (IMC) hardware accelerators for neural networks. However, most existing optimization frameworks target a single workload, leading to highly specialized hardware designs that do not generalize well across models and applications. In contrast, practical deployment scenarios require a single IMC platform that can efficiently support multiple neural network workloads. This work presents a joint hardware-workload co-optimization framework based on an optimized evolutionary algorithm for designing generalized IMC accelerator architectures. By explicitly capturing cross-workload trade-offs rather than optimizing for a single model, the proposed approach significantly reduces the performance gap between workload-specific and generalized IMC designs. The framework is evaluated on both RRAM- and SRAM-based IMC architectures, demonstrating strong robustness and adaptability across diverse design scenarios. Compared to baseline methods, the optimized designs achieve energy-delay-area product (EDAP) reductions of up to 76.2% and 95.5% when optimizing across a small set (4 workloads) and a large set (9 workloads), respectively. The source code of the framework is available at https://github.com/OlgaKrestinskaya/JointHardwareWorkloadOptimizationIMC.

Paperid: 189, https://arxiv.org/pdf/2603.03871.pdf GitHub

Abstract:
Infrared and visible image fusion (IVIF) integrates complementary modalities to enhance scene perception. Current methods predominantly focus on optimizing handcrafted losses and objective metrics, often resulting in fusion outcomes that do not align with human visual preferences. This challenge is further exacerbated by the ill-posed nature of IVIF, which severely limits its effectiveness in human perceptual environments such as security surveillance and driver assistance systems. To address these limitations, we propose a feedback reinforcement framework that bridges human evaluation to infrared and visible image fusion. To address the lack of human-centric evaluation metrics and data, we introduce the first large-scale human feedback dataset for IVIF, containing multidimensional subjective scores and artifact annotations, and enriched by a fine-tuned large language model with expert review. Based on this dataset, we design a domain-specific reward function and train a reward model to quantify perceptual quality. Guided by this reward, we fine-tune the fusion network through Group Relative Policy Optimization, achieving state-of-the-art performance that better aligns fused images with human aesthetics. Code is available at https://github.com/ALKA-Wind/EVAFusion.

Paperid: 190, https://arxiv.org/pdf/2603.03827.pdf GitHub

Abstract:
Multimodal intent recognition aims to infer human intents by jointly modeling various modalities, playing a pivotal role in real-world dialogue systems. However, current methods struggle to model hierarchical semantics underlying complex intents and lack the capacity for self-evolving reasoning over multimodal representations. To address these issues, we propose HIER, a novel method that integrates HIerarchical semantic representation with Evolutionary Reasoning based on Multimodal Large Language Model (MLLM). Inspired by human cognition, HIER introduces a structured reasoning paradigm that organizes multimodal semantics into three progressively abstracted levels. It starts with modality-specific tokens capturing localized semantic cues, which are then clustered via a label-guided strategy to form mid-level semantic concepts. To capture higher-order structure, inter-concept relations are selected using JS divergence scores to highlight salient dependencies across concepts. These hierarchical representations are then injected into MLLM via CoT-driven prompting, enabling step-wise reasoning. Besides, HIER utilizes a self-evolution mechanism that refines semantic representations through MLLM feedback, allowing dynamic adaptation during inference. Experiments on three challenging benchmarks show that HIER consistently outperforms state-of-the-art methods and MLLMs with 1-3% gains across all metrics. Code and more results are available at https://github.com/thuiar/HIER.

Paperid: 191, https://arxiv.org/pdf/2603.03825.pdf GitHub

Abstract:
The cold-start initialization stage plays a pivotal role in training Multimodal Large Reasoning Models (MLRMs), yet its mechanisms remain insufficiently understood. To analyze this stage, we introduce the Visual Attention Score (VAS), an attention-based metric that quantifies how much a model attends to visual tokens. We find that reasoning performance is strongly correlated with VAS (r=0.9616): models with higher VAS achieve substantially stronger multimodal reasoning. Surprisingly, multimodal cold-start fails to elevate VAS, resulting in attention distributions close to the base model, whereas text-only cold-start leads to a clear increase. We term this counter-intuitive phenomenon Lazy Attention Localization. To validate its causal role, we design training-free interventions that directly modulate attention allocation during inference, performance gains of 1$-$2% without any retraining. Building on these insights, we further propose Attention-Guided Visual Anchoring and Reflection (AVAR), a comprehensive cold-start framework that integrates visual-anchored data synthesis, attention-guided objectives, and visual-anchored reward shaping. Applied to Qwen2.5-VL-7B, AVAR achieves an average gain of 7.0% across 7 multimodal reasoning benchmarks. Ablation studies further confirm that each component of AVAR contributes step-wise to the overall gains. The code, data, and models are available at https://github.com/lrlbbzl/Qwen-AVAR.

Paperid: 192, https://arxiv.org/pdf/2603.03818.pdf GitHub

Abstract:
Continual learning is a long-standing challenge in robot policy learning, where a policy must acquire new skills over time without catastrophically forgetting previously learned ones. While prior work has extensively studied continual learning in relatively small behavior cloning (BC) policy models trained from scratch, its behavior in modern large-scale pretrained Vision-Language-Action (VLA) models remains underexplored. In this work, we found that pretrained VLAs are remarkably resistant to forgetting compared with smaller policy models trained from scratch. Simple Experience Replay (ER) works surprisingly well on VLAs, sometimes achieving zero forgetting even with a small replay data size. Our analysis reveals that pretraining plays a critical role in downstream continual learning performance: large pretrained models mitigate forgetting with a small replay buffer size while maintaining strong forward learning capabilities. Furthermore, we found that VLAs can retain relevant knowledge from prior tasks despite performance degradation during learning new tasks. This knowledge retention enables rapid recovery of seemingly forgotten skills through finetuning. Together, these insights imply that large-scale pretraining fundamentally changes the dynamics of continual learning, enabling models to continually acquire new skills over time with simple replay. Code and more information can be found at https://ut-austin-rpl.github.io/continual-vla

Paperid: 193, https://arxiv.org/pdf/2603.03805.pdf GitHub

Abstract:
Relational Databases (RDBs) are the backbone of modern business, yet they lack foundation models comparable to those in text or vision. A key obstacle is that high-quality RDBs are private, scarce and structurally heterogeneous, making internet-scale pre-training infeasible. To overcome this data scarcity, We introduce $\textbf{RDB-PFN}$, the first relational foundation model trained purely via $\textbf{synthetic data}$. Inspired by Prior-Data Fitted Networks (PFNs) where synthetic data generated from Structural Causal Models (SCMs) enables reasoning on single tables, we design a $\textbf{Relational Prior Generator}$ to create an infinite stream of diverse RDBs from scratch. Pre-training on $\textbf{over 2 million}$ synthetic single-table and relational tasks, RDB-PFN learns to adapt to any new database instantly via genuine $\textbf{in-context learning}$. Experiments verify RDB-PFN achieves strong few-shot performance on 19 real-world relational prediction tasks, outperforming graph-based and single-table foundation-model baselines (given the same DFS-linearized inputs), while using a lightweight architecture and fast inference. The code is available at https://github.com/MuLabPKU/RDBPFN

Paperid: 194, https://arxiv.org/pdf/2603.03796.pdf GitHub

Abstract:
When continual test-time adaptation (TTA) persists over the long term, errors accumulate in the model and further cause it to predict only a few classes for all inputs, a phenomenon known as model collapse. Recent studies have explored reset strategies that completely erase these accumulated errors. However, their periodic resets lead to suboptimal adaptation, as they occur independently of the actual risk of collapse. Moreover, their full resets cause catastrophic loss of knowledge acquired over time, even though such knowledge could be beneficial in the future. To this end, we propose (1) an Adaptive and Selective Reset (ASR) scheme that dynamically determines when and where to reset, (2) an importance-aware regularizer to recover essential knowledge lost due to reset, and (3) an on-the-fly adaptation adjustment scheme to enhance adaptability under challenging domain shifts. Extensive experiments across long-term TTA benchmarks demonstrate the effectiveness of our approach, particularly under challenging conditions. Our code is available at https://github.com/YonseiML/asr.

Paperid: 195, https://arxiv.org/pdf/2603.03790.pdf GitHub GitHub

Abstract:
Think about how human handles complex reading tasks: marking key points, inferring their relationships, and structuring information to guide understanding and responses. Likewise, can a large language model benefit from text structure to enhance text-processing performance? To explore it, in this work, we first introduce Structure of Thought (SoT), a prompting technique that explicitly guides models to construct intermediate text structures, consistently boosting performance across eight tasks and three model families. Building upon this insight, we present T2S-Bench, the first benchmark designed to evaluate and improve text-to-structure capabilities of models. T2S-Bench includes 1.8K samples across 6 scientific domains and 32 structural types, rigorously constructed to ensure accuracy, fairness, and quality. Evaluation on 45 mainstream models reveals substantial improvement potential: the average accuracy on the multi-hop reasoning task is only 52.1%, and even the most advanced model achieves 58.1% node accuracy in end-to-end extraction. Furthermore, on Qwen2.5-7B-Instruct, SoT alone yields an average +5.7% improvement across eight diverse text-processing tasks, and fine-tuning on T2S-Bench further increases this gain to +8.6%. These results highlight the value of explicit text structuring and the complementary contributions of SoT and T2S-Bench. Dataset and eval code have been released at https://t2s-bench.github.io/T2S-Bench-Page/.

Paperid: 196, https://arxiv.org/pdf/2603.03781.pdf GitHub

Abstract:
Long-term memory is fundamental for personalized agents capable of accumulating knowledge, reasoning over user experiences, and adapting across time. However, existing memory benchmarks primarily target declarative memory, specifically semantic and episodic types, where all information is explicitly presented in dialogues. In contrast, real-world actions are also governed by non-declarative memory, including habitual and procedural types, and need to be inferred from diverse digital traces. To bridge this gap, we introduce Lifebench, which features densely connected, long-horizon event simulation. It pushes AI agents beyond simple recall, requiring the integration of declarative and non-declarative memory reasoning across diverse and temporally extended contexts. Building such a benchmark presents two key challenges: ensuring data quality and scalability. We maintain data quality by employing real-world priors, including anonymized social surveys, map APIs, and holiday-integrated calendars, thus enforcing fidelity, diversity and behavioral rationality within the dataset. Towards scalability, we draw inspiration from cognitive science and structure events according to their partonomic hierarchy; enabling efficient parallel generation while maintaining global coherence. Performance results show that top-tier, state-of-the-art memory systems reach just 55.2\% accuracy, highlighting the inherent difficulty of long-horizon retrieval and multi-source integration within our proposed benchmark. The dataset and data synthesis code are available at https://github.com/1754955896/LifeBench.

Paperid: 197, https://arxiv.org/pdf/2603.03744.pdf GitHub

Abstract:
Estimating accurate, view-consistent geometry and camera poses from uncalibrated multi-view/video inputs remains challenging - especially at high spatial resolutions and over long sequences. We present DAGE, a dual-stream transformer whose main novelty is to disentangle global coherence from fine detail. A low-resolution stream operates on aggressively downsampled frames with alternating frame/global attention to build a view-consistent representation and estimate cameras efficiently, while a high-resolution stream processes the original images per-frame to preserve sharp boundaries and small structures. A lightweight adapter fuses these streams via cross-attention, injecting global context without disturbing the pretrained single-frame pathway. This design scales resolution and clip length independently, supports inputs up to 2K, and maintains practical inference cost. DAGE delivers sharp depth/pointmaps, strong cross-view consistency, and accurate poses, establishing new state-of-the-art results for video geometry estimation and multi-view reconstruction.

Paperid: 198, https://arxiv.org/pdf/2603.03726.pdf GitHub

Abstract:
No-Reference Point Cloud Quality Assessment (NR-PCQA) still struggles with generalization, primarily due to the scarcity of annotated point cloud datasets. Since the Human Visual System (HVS) drives perceptual quality assessment independently of media types, prior knowledge on quality learned from images can be repurposed for point clouds. This insight motivates adopting Unsupervised Domain Adaptation (UDA) to transfer quality-relevant priors from labeled images to unlabeled point clouds. However, existing UDA-based PCQA methods often overlook key characteristics of perceptual quality, such as sensitivity to quality ranking and quality-aware feature alignment, thereby limiting their effectiveness. To address these issues, we propose a novel Quality-aware Domain adaptation framework for PCQA, termed QD-PCQA. The framework comprises two main components: i) a Rank-weighted Conditional Alignment (RCA) strategy that aligns features under consistent quality levels and adaptively emphasizes misranked samples to reinforce perceptual quality ranking awareness; and ii) a Quality-guided Feature Augmentation (QFA) strategy, which includes quality-guided style mixup, multi-layer extension, and dual-domain augmentation modules to augment perceptual feature alignment. Extensive cross-domain experiments demonstrate that QD-PCQA significantly improves generalization in NR-PCQA tasks. The code is available at https://github.com/huhu-code/QD-PCQA.

Paperid: 199, https://arxiv.org/pdf/2603.03718.pdf GitHub

Abstract:
Glass surface segmentation from RGB images is a challenging task, since glass as a transparent material distinctly lacks visual characteristics. However, glass segmentation is critical for scene understanding and robotics, as transparent glass surfaces must be identified as solid material. This paper presents a novel architecture for glass segmentation, deploying a dual-backbone producing general visual features as well as task-specific learned visual features. General visual features are produced by a frozen DINOv3 vision foundation model, and the task-specific features are generated with a Swin model trained in a supervised manner. Resulting multi-scale feature representations are downsampled with residual Squeeze-and-Excitation Channel Reduction, and fed into a Mask2Former Decoder, producing the final segmentation masks. The architecture was evaluated on four commonly used glass segmentation datasets, achieving state-of-the-art results on several accuracy metrics. The model also has a competitive inference speed compared to the previous state-of-the-art method, and surpasses it when using a lighter DINOv3 backbone variant. The implementation source code and model weights are available at: https://github.com/ojalar/lgnet

Paperid: 200, https://arxiv.org/pdf/2603.03692.pdf GitHub

Abstract:
Classifier-Free Guidance (CFG) has established the foundation for guidance mechanisms in diffusion models, showing that well-designed guidance proxies significantly improve conditional generation and sample quality. Autoguidance (AG) has extended this idea, but it relies on an auxiliary network and leaves solver-induced errors unaddressed. In stiff regions, the ODE trajectory changes sharply, where local truncation error (LTE) becomes a critical factor that deteriorates sample quality. Our key observation is that these errors align with the dominant eigenvector, motivating us to leverage the solver-induced error as a guidance signal. We propose Embedded Runge-Kutta Guidance (ERK-Guid), which exploits detected stiffness to reduce LTE and stabilize sampling. We theoretically and empirically analyze stiffness and eigenvector estimators with solver errors to motivate the design of ERK-Guid. Our experiments on both synthetic datasets and the popular benchmark dataset, ImageNet, demonstrate that ERK-Guid consistently outperforms state-of-the-art methods. Code is available at https://github.com/mlvlab/ERK-Guid.

Paperid: 201, https://arxiv.org/pdf/2603.03680.pdf GitHub

Abstract:
Large Language Model (LLM) agents have demonstrated remarkable proficiency in learned tasks, yet they often struggle to adapt to non-stationary environments with feedback. While In-Context Learning and external memory offer some flexibility, they fail to internalize the adaptive ability required for long-term improvement. Meta-Reinforcement Learning (meta-RL) provides an alternative by embedding the learning process directly within the model. However, existing meta-RL approaches for LLMs focus primarily on exploration in single-agent settings, neglecting the strategic exploitation necessary for multi-agent environments. We propose MAGE, a meta-RL framework that empowers LLM agents for strategic exploration and exploitation. MAGE utilizes a multi-episode training regime where interaction histories and reflections are integrated into the context window. By using the final episode reward as the objective, MAGE incentivizes the agent to refine its strategy based on past experiences. We further combine population-based training with an agent-specific advantage normalization technique to enrich agent diversity and ensure stable learning. Experiment results show that MAGE outperforms existing baselines in both exploration and exploitation tasks. Furthermore, MAGE exhibits strong generalization to unseen opponents, suggesting it has internalized the ability for strategic exploration and exploitation. Code is available at https://github.com/Lu-Yang666/MAGE.

Paperid: 202, https://arxiv.org/pdf/2603.03657.pdf GitHub

Abstract:
Multimodal generative models have made significant strides in image editing, demonstrating impressive performance on a variety of static tasks. However, their proficiency typically does not extend to complex scenarios requiring dynamic reasoning, leaving them ill-equipped to model the coherent, intermediate logical pathways that constitute a multi-step evolution from an initial state to a final one. This capacity is crucial for unlocking a deeper level of procedural and causal understanding in visual manipulation. To systematically measure this critical limitation, we introduce InEdit-Bench, the first evaluation benchmark dedicated to reasoning over intermediate pathways in image editing. InEdit-Bench comprises meticulously annotated test cases covering four fundamental task categories: state transition, dynamic process, temporal sequence, and scientific simulation. Additionally, to enable fine-grained evaluation, we propose a set of assessment criteria to evaluate the logical coherence and visual naturalness of the generated pathways, as well as the model's fidelity to specified path constraints. Our comprehensive evaluation of 14 representative image editing models on InEdit-Bench reveals significant and widespread shortcomings in this domain. By providing a standardized and challenging benchmark, we aim for InEdit-Bench to catalyze research and steer development towards more dynamic, reason-aware, and intelligent multimodal generative models.

Paperid: 203, https://arxiv.org/pdf/2603.03617.pdf GitHub

Abstract:
RGB-Thermal (RGBT) tracking aims to achieve robust object localization across diverse environmental conditions by fusing visible and thermal infrared modalities. However, existing RGBT trackers rely solely on initial-frame visual information for target modeling, failing to adapt to appearance variations due to the absence of language guidance. Furthermore, current methods suffer from redundant search regions and heterogeneous modality gaps, causing background distraction. To address these issues, we first introduce textual descriptions into RGBT tracking benchmarks. This is accomplished through a pipeline that leverages Multi-modal Large Language Models (MLLMs) to automatically produce texual annotations. Afterwards, we propose RAGTrack, a novel Retrieval-Augmented Generation framework for robust RGBT tracking. To this end, we introduce a Multi-modal Transformer Encoder (MTE) for unified visual-language modeling. Then, we design an Adaptive Token Fusion (ATF) to select target-relevant tokens and perform channel exchanges based on cross-modal correlations, mitigating search redundancies and modality gaps. Finally, we propose a Context-aware Reasoning Module (CRM) to maintain a dynamic knowledge base and employ a Retrieval-Augmented Generation (RAG) to enable temporal linguistic reasoning for robust target modeling. Extensive experiments on four RGBT benchmarks demonstrate that our framework achieves state-of-the-art performance across various challenging scenarios. The source code is available https://github.com/IdolLab/RAGTrack.

Paperid: 204, https://arxiv.org/pdf/2603.03602.pdf GitHub

Abstract:
The automatic design of a 3D tooth model plays a crucial role in dental digitization. However, current approaches face challenges in compositional 3D tooth generation because both the layouts and shapes of missing teeth need to be optimized.In addition, collision conflicts are often omitted in 3D Gaussian-based compositional 3D generation, where objects may intersect with each other due to the absence of explicit geometric information on the object surfaces. Motivated by graph generation through diffusion models and collision detection using 3D Gaussians, we propose an approach named DM-CFO for compositional tooth generation, where the layout of missing teeth is progressively restored during the denoising phase under both text and graph constraints. Then, the Gaussian parameters of each layout-guided tooth and the entire jaw are alternately updated using score distillation sampling (SDS). Furthermore, a regularization term based on the distances between the 3D Gaussians of neighboring teeth and the anchor tooth is introduced to penalize tooth intersections. Experimental results on three tooth-design datasets demonstrate that our approach significantly improves the multiview consistency and realism of the generated teeth compared with existing methods. Project page: https://amateurc.github.io/CF-3DTeeth/.

Paperid: 205, https://arxiv.org/pdf/2603.03587.pdf GitHub

Abstract:
Method validation and study design in causal inference rely on synthetic data with known counterfactuals. Existing simulators trade off distributional realism, the ability to capture mixed-type and multimodal tabular data, against causal controllability, including explicit control over overlap, unmeasured confounding, and treatment effect heterogeneity. We introduce CausalMix, a variational generative framework that closes this gap by coupling a mixture of Gaussian latent priors with data-type-specific decoders for continuous, binary, and categorical variables. The model incorporates explicit causal controls: an overlap regularizer shaping propensity-score distributions, alongside direct parameterizations of confounding strength and effect heterogeneity. This unified objective preserves fidelity to the observed data while enabling factorial manipulation of causal mechanisms, allowing overlap, confounding strength, and treatment effect heterogeneity to be varied independently at design time. Across benchmarks, CausalMix achieves state-of-the-art distributional metrics on mixed-type tables while providing stable, fine-grained causal control. We demonstrate practical utility in a comparative safety study of metastatic castration-resistant prostate cancer treatments, using CausalMix to compare estimators under calibrated data-generating processes, tune hyperparameters, and conduct simulation-based power analyses under targeted treatment effect heterogeneity scenarios.

Paperid: 206, https://arxiv.org/pdf/2603.03580.pdf GitHub

Abstract:
Scene text recognition (STR) and handwritten text recognition (HTR) face significant challenges in accurately transcribing textual content from images into machine-readable formats. Conventional OCR models often predict transcriptions directly, which limits detailed reasoning about text structure. We propose a VQA-inspired data augmentation framework that strengthens OCR training through structured question-answering tasks. For each image-text pair, we generate natural-language questions probing character-level attributes such as presence, position, and frequency, with answers derived from ground-truth text. These auxiliary tasks encourage finer-grained reasoning, and the OCR model aligns visual features with textual queries to jointly reason over images and questions. Experiments on WordArt and Esposalles datasets show consistent improvements over baseline models, with significant reductions in both CER and WER. Our code is publicly available at https://github.com/xuyaooo/DataAugOCR.

Paperid: 207, https://arxiv.org/pdf/2603.03573.pdf GitHub

Abstract:
Discrete biological sequence optimization requires iterative refinement under strict syntactic constraints. Diffusion models offer progressive refinement but do not naturally expose controllable discrete edit operations, while autoregressive LLMs often lack explicit long-horizon planning for constrained edits. We propose STRIDE (Sequence Trajectory Refinement via Internalized Denoising Emulation), a post-training framework that trains an LLM to emit executable trajectories of atomic edits (INSERT/DELETE/REPLACE) as a verifiable reasoning trace for variable-length refinement. STRIDE combines supervised fine-tuning on Levenshtein-aligned shortest edit demonstrations with group-based policy optimization to align edit trajectories with task rewards while preserving coherent editing behavior. Across protein fluorescence and instruction-conditioned molecular optimization, STRIDE improves variable-length protein editing success from 42% to 89% while increasing novelty from 47% to 97%, and yields stronger validity and controllability compared to diverse baselines. The code is published at https://github.com/daiheng-zhang/STRIDE.

Paperid: 208, https://arxiv.org/pdf/2603.03530.pdf GitHub GitHub

Abstract:
Frozen self-supervised representations often transfer well with only a few labels across many semantic tasks. We argue that a single geometric quantity, \emph{directional} CDNV (decision-axis variance), sits at the core of two favorable behaviors: strong few-shot transfer within a task, and low interference across many tasks. We show that both emerge when variability \emph{along} class-separating directions is small. First, we prove sharp non-asymptotic multiclass generalization bounds for downstream classification whose leading term is the directional CDNV. The bounds include finite-shot corrections that cleanly separate intrinsic decision-axis variability from centroid-estimation error. Second, we link decision-axis collapse to multitask geometry: for independent balanced labelings, small directional CDNV across tasks forces the corresponding decision axes to be nearly orthogonal, helping a single representation support many tasks with minimal interference. Empirically, across SSL objectives, directional CDNV collapses during pretraining even when classical CDNV remains large, and our bounds closely track few-shot error at practical shot sizes. Additionally, on synthetic multitask data, we verify that SSL learns representations whose induced decision axes are nearly orthogonal. The code and project page of the paper are available at [\href{https://dlfundamentals.github.io/directional-neural-collapse/}{project page}].

Paperid: 209, https://arxiv.org/pdf/2603.03529.pdf GitHub

Abstract:
We introduce mlx-snn, the first spiking neural network (SNN) library built natively on Apple's MLX framework. As SNN research grows rapidly, all major libraries -- snnTorch, Norse, SpikingJelly, Lava -- target PyTorch or custom backends, leaving Apple Silicon users without a native option. mlx-snn provides six neuron models (LIF, IF, Izhikevich, Adaptive LIF, Synaptic, Alpha), four surrogate gradient functions, four spike encoding methods (including an EEG-specific encoder), and a complete backpropagation-through-time training pipeline. The library leverages MLX's unified memory architecture, lazy evaluation, and composable function transforms (mx.grad, mx.compile) to enable efficient SNN research on Apple Silicon hardware. We validate mlx-snn on MNIST digit classification across five hyperparameter configurations and three backends, achieving up to 97.28% accuracy with 2.0--2.5 times faster training and 3--10 times lower GPU memory than snnTorch on the same M3 Max hardware. mlx-snn is open-source under the MIT license and available on PyPI. https://github.com/D-ST-Sword/mlx-snn

Paperid: 210, https://arxiv.org/pdf/2603.03482.pdf GitHub

Abstract:
Interactive world models continually generate video by responding to a user's actions, enabling open-ended generation capabilities. However, existing models typically lack a 3D representation of the environment, meaning 3D consistency must be implicitly learned from data, and spatial memory is restricted to limited temporal context windows. This results in an unrealistic user experience and presents significant obstacles to down-stream tasks such as training agents. To address this, we present PERSIST, a new paradigm of world model which simulates the evolution of a latent 3D scene: environment, camera, and renderer. This allows us to synthesize new frames with persistent spatial memory and consistent geometry. Both quantitative metrics and a qualitative user study show substantial improvements in spatial memory, 3D consistency, and long-horizon stability over existing methods, enabling coherent, evolving 3D worlds. We further demonstrate novel capabilities, including synthesising diverse 3D environments from a single image, as well as enabling fine-grained, geometry-aware control over generated experiences by supporting environment editing and specification directly in 3D space. Project page: https://francelico.github.io/persist.github.io

Paperid: 211, https://arxiv.org/pdf/2603.03454.pdf GitHub

Abstract:
Offline Reinforcement Learning (RL) is an emerging field of RL in which policies are learned solely from demonstrations. Within offline RL, some environments involve balancing multiple objectives, but existing multi-objective offline RL algorithms do not provide an efficient way to find a fair compromise. FairDICE (see arXiv:2506.08062v2) seeks to fill this gap by adapting OptiDICE (an offline RL algorithm) to automatically learn weights for multiple objectives to e.g.\ incentivise fairness among objectives. As this would be a valuable contribution, this replication study examines the replicability of claims made regarding FairDICE. We find that many theoretical claims hold, but an error in the code reduces FairDICE to standard behaviour cloning in continuous environments, and many important hyperparameters were originally underspecified. After rectifying this, we show in experiments extending the original paper that FairDICE can scale to complex environments and high-dimensional rewards, though it can be reliant on (online) hyperparameter tuning. We conclude that FairDICE is a theoretically interesting method, but the experimental justification requires significant revision.

Paperid: 212, https://arxiv.org/pdf/2603.03418.pdf GitHub

Abstract:
Recently, DeepSeek has invented the manifold-constrained hyper-connection (mHC) approach which has demonstrated significant improvements over the traditional residual connection in deep learning models \cite{xie2026mhc}. Nevertheless, this approach has not been tailor-designed for improving hyperspectral image (HSI) classification. This paper presents a clustering-guided mHC Mamba model (mHC-HSI) for enhanced HSI classification, with the following contributions. First, to improve spatial-spectral feature learning, we design a novel clustering-guided Mamba module, based on the mHC framework, that explicitly learns both spatial and spectral information in HSI. Second, to decompose the complex and heterogeneous HSI into smaller clusters, we design a new implementation of the residual matrix in mHC, which can be treated as soft cluster membership maps, leading to improved explainability of the mHC approach. Third, to leverage the physical spectral knowledge, we divide the spectral bands into physically-meaningful groups and use them as the "parallel streams" in mHC, leading to a physically-meaningful approach with enhanced interpretability. The proposed approach is tested on benchmark datasets in comparison with the state-of-the-art methods, and the results suggest that the proposed model not only improves the accuracy but also enhances the model explainability. Code is available here: https://github.com/GSIL-UCalgary/mHC_HyperSpectral

Paperid: 213, https://arxiv.org/pdf/2603.03415.pdf GitHub

Abstract:
In this work, we investigate how Large Language Models (LLMs) adapt their internal representations when encountering inputs of increasing difficulty, quantified as the degree of out-of-distribution (OOD) shift. We reveal a consistent and quantifiable phenomenon: as task difficulty increases, whether through harder reasoning questions, longer contexts, or adding answer choices, the last hidden states of LLMs become substantially sparser. In short, \textbf{\textit{the farther the shift, the sparser the representations}}. This sparsity--difficulty relation is observable across diverse models and domains, suggesting that language models respond to unfamiliar or complex inputs by concentrating computation into specialized subspaces in the last hidden state. Through a series of controlled analyses with a learning dynamic explanation, we demonstrate that this sparsity is not incidental but an adaptive mechanism for stabilizing reasoning under OOD. Leveraging this insight, we design \textit{Sparsity-Guided Curriculum In-Context Learning (SG-ICL)}, a strategy that explicitly uses representation sparsity to schedule few-shot demonstrations, leading to considerable performance enhancements. Our study provides new mechanistic insights into how LLMs internalize OOD challenges. The source code is available at the URL: https://github.com/MingyuJ666/sparsityLLM.

Paperid: 214, https://arxiv.org/pdf/2603.03412.pdf GitHub

Abstract:
Recent advances in generative image editing have enabled transformative applications, from professional head shot generation to avatar stylization. However, these systems often require uploading high-fidelity facial images to third-party models, raising concerns around biometric privacy, data misuse, and user consent. We propose a privacy-preserving pipeline that supports high-quality editing while keeping users in control over their biometric data in face-centric use cases. Our approach separates identity-sensitive regions from editable image context using on-device segmentation and masking, enabling secure, user-controlled editing without modifying third-party generative models. Unlike traditional cloud-based tools, PRIVATEEDIT enforces privacy by default: biometric data is never exposed or transmitted. This design requires no access to or retraining of third-party models, making it compatible with a wide range of commercial APIs. By treating privacy as a core design constraint, our system supports responsible generative AI centered on user autonomy and trust. The pipeline includes a tunable masking mechanism that lets users control how much facial information is concealed, allowing them to balance privacy and output fidelity based on trust level or use case. We demonstrate its applicability in professional and creative workflows and provide a user interface for selective anonymization. By advocating privacy-by-design in generative AI, our work offers both technical feasibility and normative guidance for protecting digital identity. The source code is available at https://github.com/Dipeshtamboli/PrivateEdit-Privacy-Preserving-GenAI.

Paperid: 215, https://arxiv.org/pdf/2603.03406.pdf GitHub

Abstract:
How should two language models interact to produce better code than either can alone? The conventional approach -- a reasoning model plans, a code specialist implements -- seems natural but fails: on HumanEval+, plan-then-code degrades performance by 2.4 percentage points versus the code specialist alone. We show that reversing the interaction changes everything. When the code specialist generates freely and the reasoning model reviews instead of plans, the same two models on the same hardware achieve 90.2% pass@1 -- exceeding GPT-4o (87.2%) and O1 Preview (89.0%) -- on ~$2/hr of commodity GPU. Cross-benchmark validation across 542 problems (HumanEval+ and MBPP+) reveals a moderating variable: review effectiveness scales with specification richness, yielding 4x more improvement on richly-specified problems (+9.8pp) than on lean ones (+2.3pp), while remaining net-positive in both cases. The practical implication is twofold: compose models by their cognitive strengths (reviewers review, coders code), and invest in specification quality to amplify the returns.

Paperid: 216, https://arxiv.org/pdf/2603.03389.pdf GitHub

Abstract:
Obtaining a single-vector representation from a Large Language Model's (LLM) token-level outputs is a critical step for nearly all sentence-level tasks. However, standard pooling methods like mean or max aggregation treat tokens as an independent set, discarding the rich relational structure captured by the model's self-attention layers and making them susceptible to signal dilution. To address this, we introduce GLOT, a lightweight, structure-aware pooling module that reframes pooling as relational learning followed by aggregation. Operating on the outputs of a frozen LLM, GLOT first constructs a latent token-similarity graph, then refines token representations with a graph neural network, and finally aggregates them using a readout layer. Experimentally, our approach is remarkably robust and efficient: on a diagnostic stress test where 90% of tokens are random distractors, GLOT maintains over 97% accuracy while baseline methods collapse. Furthermore, it is competitive with state-of-the-art techniques on benchmarks like GLUE and MTEB with 20x fewer trainable parameters and speeds up the training time by over 100x compared with parameter-efficient fine-tuning methods. Supported by a theoretical analysis of its expressive power, our work shows that learning over token graphs is a powerful paradigm for the efficient adaptation of frozen LLMs. Our code is published at https://github.com/ipsitmantri/GLOT.

Paperid: 217, https://arxiv.org/pdf/2603.03379.pdf GitHub

Abstract:
As Large Language Models (LLMs) are increasingly used for long-duration tasks, maintaining effective long-term memory has become a critical challenge. Current methods often face a trade-off between cost and accuracy. Simple storage methods often fail to retrieve relevant information, while complex indexing methods (such as memory graphs) require heavy computation and can cause information loss. Furthermore, relying on the working LLM to process all memories is computationally expensive and slow. To address these limitations, we propose MemSifter, a novel framework that offloads the memory retrieval process to a small-scale proxy model. Instead of increasing the burden on the primary working LLM, MemSifter uses a smaller model to reason about the task before retrieving the necessary information. This approach requires no heavy computation during the indexing phase and adds minimal overhead during inference. To optimize the proxy model, we introduce a memory-specific Reinforcement Learning (RL) training paradigm. We design a task-outcome-oriented reward based on the working LLM's actual performance in completing the task. The reward measures the actual contribution of retrieved memories by mutiple interactions with the working LLM, and discriminates retrieved rankings by stepped decreasing contributions. Additionally, we employ training techniques such as Curriculum Learning and Model Merging to improve performance. We evaluated MemSifter on eight LLM memory benchmarks, including Deep Research tasks. The results demonstrate that our method meets or exceeds the performance of existing state-of-the-art approaches in both retrieval accuracy and final task completion. MemSifter offers an efficient and scalable solution for long-term LLM memory. We have open-sourced the model weights, code, and training data to support further research.

Paperid: 218, https://arxiv.org/pdf/2603.03372.pdf GitHub

Abstract:
Density Functional Theory (DFT) is a cornerstone of materials science, yet executing DFT in practice requires coordinating a complex, multi-step workflow. Existing tools and LLM-based solutions automate parts of the steps, but lack support for full workflow automation, diverse task adaptation, and accuracy-cost trade-off optimization in DFT configuration. To this end, we present TritonDFT, a multi-agent framework that enables efficient and accurate DFT execution through an expert-curated, extensible workflow design, Pareto-aware parameter inference, and multi-source knowledge augmentation. We further introduce DFTBench, a benchmark for evaluating the agent's multi-dimensional capabilities, spanning science expertise, trade0off optimization, HPC knowledge, and cost efficiency. TritonDFT provides an open user interface for real-world usage. Our website is at https://www.tritondft.com. Our source code and benchmark suite are available at https://github.com/Leo9660/TritonDFT.git.

Paperid: 219, https://arxiv.org/pdf/2603.03354.pdf GitHub

Abstract:
Although obtaining deep brain activity from non-invasive scalp electroencephalography (sEEG) is crucial for neuroscience and clinical diagnosis, directly generating high-fidelity intracranial electroencephalography (iEEG) signals remains a largely unexplored field, limiting our understanding of deep brain dynamics. Current research primarily focuses on traditional signal processing or source localization methods, which struggle to capture the complex waveforms and random characteristics of iEEG. To address this critical challenge, this paper introduces NeuroFlowNet, a novel cross-modal generative framework whose core contribution lies in the first-ever reconstruction of iEEG signals from the entire deep temporal lobe region using sEEG signals. NeuroFlowNet is built on Conditional Normalizing Flow (CNF), which directly models complex conditional probability distributions through reversible transformations, thereby explicitly capturing the randomness of brain signals and fundamentally avoiding the pattern collapse issues common in existing generative models. Additionally, the model integrates a multi-scale architecture and self-attention mechanisms to robustly capture fine-grained temporal details and long-range dependencies. Validation results on a publicly available synchronized sEEG-iEEG dataset demonstrate NeuroFlowNet's effectiveness in terms of temporal waveform fidelity, spectral feature reproduction, and functional connectivity restoration. This study establishes a more reliable and scalable new paradigm for non-invasive analysis of deep brain dynamics. The code of this study is available in https://github.com/hdy6438/NeuroFlowNet

Paperid: 220, https://arxiv.org/pdf/2603.03335.pdf GitHub

Abstract:
Large language models (LLMs) exhibit a wide range of capabilities, including mathematical reasoning, code generation, and linguistic behaviors. We show that many capabilities are highly localized to small subsets of attention heads within Transformer architectures. Zeroing out as few as five task-specific heads can degrade performance by up to $65\%$ on standard benchmarks measuring the capability of interest, while largely preserving performance on unrelated tasks. We introduce a compressed sensing based method that exploits the sparsity of these heads to identify them via strategic knockouts and a small number of model evaluations. We validate these findings across Llama and Qwen models ranging from 1B to 8B parameters and a diverse set of capabilities including mathematical abilities and code generation, revealing a modular organization in which specialized capabilities are implemented by sparse, functionally distinct components. Overall, our results suggest that capability localization is a general organizational principle of Transformer language models, with implications for interpretability, model editing, and AI safety. Code is released at https://github.com/locuslab/llm-components.

Paperid: 221, https://arxiv.org/pdf/2603.03334.pdf GitHub

Abstract:
The evaluation of Large Language Models (LLMs) on mathematical reasoning has largely focused on elementary problems, competition-style questions, or formal theorem proving, leaving graduate-level and computational mathematics relatively underexplored. We introduce CompMath-MCQ, a new benchmark dataset for assessing LLMs on advanced mathematical reasoning in a multiple-choice setting. The dataset consists of 1{,}500 originally authored questions by professors of graduate-level courses, covering topics including Linear Algebra, Numerical Optimization, Vector Calculus, Probability, and Python-based scientific computing. Three option choices are provided for each question, with exactly one of them being correct. To ensure the absence of data leakage, all questions are newly created and not sourced from existing materials. The validity of questions is verified through a procedure based on cross-LLM disagreement, followed by manual expert review. By adopting a multiple-choice format, our dataset enables objective, reproducible, and bias-free evaluation through lm_eval library. Baseline results with state-of-the-art LLMs indicate that advanced computational mathematical reasoning remains a significant challenge. We release CompMath-MCQ at the following link: https://github.com/biancaraimondi/CompMath-MCQ.git

Paperid: 222, https://arxiv.org/pdf/2603.03332.pdf GitHub

Abstract:
Chain-of-Thought (CoT) prompting has emerged as a foundational technique for eliciting reasoning from Large Language Models (LLMs), yet the robustness of this approach to corruptions in intermediate reasoning steps remains poorly understood. This paper presents a comprehensive empirical evaluation of LLM robustness to a structured taxonomy of 5 CoT perturbation types: \textit{MathError, UnitConversion, Sycophancy, SkippedSteps,} and \textit{ExtraSteps}. We evaluate 13 models spanning three orders of magnitude in parameter count (3B to 1.5T\footnote{Assumed parameter count of closed models}), testing their ability to complete mathematical reasoning tasks despite perturbations injected at different points in the reasoning chain. Our key findings reveal heterogeneous vulnerability patterns: MathError perturbations produce the most severe degradation in small models (50-60\% accuracy loss) but show strong scaling benefits; UnitConversion remains challenging across all scales (20-30\% loss even for largest models); ExtraSteps incur minimal accuracy degradation (0-6\%) regardless of scale; Sycophancy produces modest effects (7\% loss for small models); and SkippedSteps cause intermediate damage (15\% loss). Scaling relationships follow power-law patterns, with model size serving as a protective factor against some perturbations but offering limited defense against dimensional reasoning tasks. These findings have direct implications for deploying LLMs in multi-stage reasoning pipelines and underscore the necessity of task-specific robustness assessments and mitigation strategies. The code and results are available https://github.com/Mystic-Slice/CoTPerturbation.

Paperid: 223, https://arxiv.org/pdf/2603.03331.pdf GitHub

Abstract:
Photoplethysmography (PPG) is a widely used non-invasive sensing modality for continuous cardiovascular and physiological monitoring across clinical, laboratory, and wearable settings. While existing PPG datasets support a broad range of downstream tasks, they typically provide supervision in the form of numerical measurements or task-specific labels, limiting their suitability for language-based physiological reasoning and multimodal foundation models. In this work, we introduce PulseLM, a large-scale PPG-text dataset designed to bridge raw PPG waveforms and natural language through a unified, closed-ended question answering (QA) formulation. PulseLM aggregates PPG recordings from fifteen publicly available sources and harmonizes heterogeneous annotations into twelve common physiologically QA tasks. The dataset comprises 1.31 million standardized 10-second PPG segments, associated with 3.15 million question-answer pairs. We further define reproducible preprocessing, supervision, and evaluation protocols and establish baseline benchmarks using multimodal PPG-aware large language models. PulseLM provides a standardized foundation for studying multimodal physiological reasoning, cross-dataset generalization, and scalable benchmarking of PPG-based language models. The data and code can be found publicly available at: https://github.com/manhph2211/PulseLM.

Paperid: 224, https://arxiv.org/pdf/2603.03328.pdf GitHub

Abstract:
Language exhibits inherent structures, a property that explains both language acquisition and language change. Given this characteristic, we expect language models to manifest internal structures as well. While interpretability research has investigated the components of language models, existing approaches focus on local inter-token relationships within layers or modules (e.g., Multi-Head Attention), leaving global inter-layer relationships largely overlooked. To address this gap, we introduce StructLens, an analytical framework designed to reveal how internal structures relate holistically through their inter-token connection within a layer. StructLens constructs maximum spanning trees based on the semantic representations in residual streams, analogous to dependency parsing, and leverages the tree properties to quantify inter-layer distance (or similarity) from a structural perspective. Our findings demonstrate that StructLens yields an inter-layer similarity pattern that is distinctively different from conventional cosine similarity. Moreover, this structure-aware similarity proves to be beneficial for practical tasks, such as layer pruning, highlighting the effectiveness of structural analysis for understanding and optimizing language models. Our code is available at https://github.com/naist-nlp/structlens.

Paperid: 225, https://arxiv.org/pdf/2603.03314.pdf GitHub

Abstract:
Large language models (LLMs) have demonstrated remarkable and steadily improving performance across a wide range of tasks. However, LLM performance may be highly sensitive to prompt variations especially in scenarios with limited openness or strict output formatting requirements, indicating insufficient robustness. In real-world applications, user prompts provided to LLMs often contain imperfections, which may undermine the quality of the model's responses. To address this issue, previous work has primarily focused on preprocessing prompts, employing external tools or even LLMs to refine prompt formulations in advance. However, these approaches overlook the intrinsic robustness of LLMs, and their reliance on external components introduces additional computational overhead and uncertainty. In this work, we propose a Contrastive Learning-based Inverse Direct Preference Optimization (CoIPO) method that minimizes the discrepancy between the label-aligned logits produced by the model under a clean prompt and its noisy counterpart, and conduct a detailed analysis using mutual information theory. We augment the FLAN dataset by constructing paired prompts, each consisting of a clean prompt and its corresponding noisy version for training. Additionally, to evaluate the effectiveness, we develop NoisyPromptBench, a benchmark enhanced and derived from the existing PromptBench. Experimental results conducted on NoisyPromptBench demonstrate that our proposed method achieves a significant improvement in average accuracy over the current state-of-the-art approaches. The source code of CoIPO, pair-wise FLAN datasets, and NoisyPromptBench have already been released on https://github.com/vegetable-yx/CoIPO.

Paperid: 226, https://arxiv.org/pdf/2603.03312.pdf GitHub

Abstract:
Decoding natural language from non-invasive EEG signals is a promising yet challenging task. However, current state-of-the-art models remain constrained by three fundamental limitations: Semantic Bias (mode collapse into generic templates), Signal Neglect (hallucination based on linguistic priors rather than neural inputs), and the BLEU Trap, where evaluation metrics are artificially inflated by high-frequency stopwords, masking a lack of true semantic fidelity. To address these challenges, we propose SemKey, a novel multi-stage framework that enforces signal-grounded generation through four decoupled semantic objectives: sentiment, topic, length, and surprisal. We redesign the interaction between the neural encoder and the Large Language Model (LLM) by injecting semantic prompts as Queries and EEG embeddings as Key-Value pairs, strictly forcing the model to attend to neural inputs. Furthermore, we move beyond standard translation metrics by adopting N-way Retrieval Accuracy and Fréchet Distance to rigorously assess diversity and alignment. Extensive experiments demonstrate that our approach effectively eliminates hallucinations on noise inputs and achieves SOTA performance on these robust protocols. Code will be released upon acceptance at https://github.com/xmed-lab/SemKey.

Paperid: 227, https://arxiv.org/pdf/2603.03308.pdf GitHub

Abstract:
How does the conversational past of large language models (LLMs) influence their future performance? Recent work suggests that LLMs are affected by their conversational history in unexpected ways. For instance, hallucinations in prior interactions may influence subsequent model responses. In this work, we introduce History-Echoes, a framework that investigates how conversational history biases subsequent generations. The framework explores this bias from two perspectives: probabilistically, we model conversations as Markov chains to quantify state consistency; geometrically, we measure the consistency of consecutive hidden representations. Across three model families and six datasets spanning diverse phenomena, our analysis reveals a strong correlation between the two perspectives. By bridging these perspectives, we demonstrate that behavioral persistence manifests as a geometric trap, where gaps in the latent space confine the model's trajectory. Code available at https://github.com/technion-cs-nlp/OldHabitsDieHard.

Paperid: 228, https://arxiv.org/pdf/2603.03306.pdf GitHub

Abstract:
Recently presented Token-Oriented Object Notation (TOON) aims to replace JSON as a serialization format for passing structured data to LLMs with significantly reduced token usage. While showing solid accuracy in LLM comprehension, there is a lack of tests against JSON generation. Though never present in training data, TOON syntax is simple enough to suggest one-shot in-context learning could support accurate generation. The inevitable prompt overhead can be an acceptable trade-off for shorter completions. To test this, we conducted a benchmark creating several test cases with regard to structural complexity, a validation pipeline, and comparing plain JSON generation vs structured output (via constrained decoding) JSON generation vs TOON one-shot in-context learning generation. JSON structured output was included to establish a minimum token budget baseline and to set a starting point for future experiments testing TOON constrained decoding inference enforcement. Key findings: TOON shows promising accuracy/token consumption ratio for in-domain generation tasks, though this advantage is often reduced by the "prompt tax" of instructional overhead in shorter contexts. Plain JSON generation shows the best one-shot and final accuracy, even compared with constrained decoding structured output, where the only significant advantage is the lowest token usage as a trade-off for slightly decreased accuracy overall and significant degradation for some models. Notably, for simple structures, this "lowest token usage" of constrained decoding outperformed even TOON, hinting that TOON enforcing via frameworks such as xgrammar may not yield the desired results. Furthermore, the results suggest a scaling hypothesis: TOON's true efficiency potential likely follows a non-linear curve, shining only beyond a specific point where cumulative syntax savings amortize the initial prompt overhead.

Paperid: 229, https://arxiv.org/pdf/2603.03298.pdf GitHub

Abstract:
Large Language Models (LLMs) have improved substantially alignment, yet their behavior remains highly sensitive to prompt phrasing. This brittleness has motivated automated prompt engineering, but most existing methods (i) require a task-specific training set, (ii) rely on expensive iterative optimization to produce a single dataset-level prompt, and (iii) must be rerun from scratch for each new task. We introduce TATRA, a dataset-free prompting method that constructs instance-specific few-shot prompts by synthesizing on-the-fly examples to accompany a user-provided instruction. TATRA requires no labeled training data and avoids task-specific optimization loops, while retaining the benefits of demonstration-based prompting. Across standard text classification benchmarks, TATRA matches or improves over strong prompt-optimization baselines that depend on training data and extensive search. On mathematical reasoning benchmarks, TATRA achieves state-of-the-art performance on GSM8K and DeepMath, outperforming methods that explicitly optimize prompts on those tasks. Our results suggest that per-instance construction of effective in-context examples is more important than running long, expensive optimization loops to produce a single prompt per task. We will make all code publicly available upon acceptance of the paper. Code is available at https://github.com/BMD223/TATRA

Paperid: 230, https://arxiv.org/pdf/2603.03296.pdf GitHub

Abstract:
Long-term memory is essential for large language model (LLM) agents operating in complex environments, yet existing memory designs are either task-specific and non-transferable, or task-agnostic but less effective due to low task-relevance and context explosion from raw memory retrieval. We propose PlugMem, a task-agnostic plugin memory module that can be attached to arbitrary LLM agents without task-specific redesign. Motivated by the fact that decision-relevant information is concentrated as abstract knowledge rather than raw experience, we draw on cognitive science to structure episodic memories into a compact, extensible knowledge-centric memory graph that explicitly represents propositional and prescriptive knowledge. This representation enables efficient memory retrieval and reasoning over task-relevant knowledge, rather than verbose raw trajectories, and departs from other graph-based methods like GraphRAG by treating knowledge as the unit of memory access and organization instead of entities or text chunks. We evaluate PlugMem unchanged across three heterogeneous benchmarks (long-horizon conversational question answering, multi-hop knowledge retrieval, and web agent tasks). The results show that PlugMem consistently outperforms task-agnostic baselines and exceeds task-specific memory designs, while also achieving the highest information density under a unified information-theoretic analysis. Code and data are available at https://github.com/TIMAN-group/PlugMem.

Paperid: 231, https://arxiv.org/pdf/2603.03292.pdf GitHub

Abstract:
Large Language Models (LLMs) exhibit high reasoning capacity in medical question-answering, but their tendency to produce hallucinations and outdated knowledge poses critical risks in healthcare fields. While Retrieval-Augmented Generation (RAG) mitigates these issues, existing methods rely on noisy token-level signals and lack the multi-round refinement required for complex reasoning. In the paper, we propose **MA-RAG** (**M**ulti-Round **A**gentic RAG), a framework that facilitates test-time scaling for complex medical reasoning by iteratively evolving both external evidence and internal reasoning history within an agentic refinement loop. At each round, the agent transforms semantic **conflict** among candidate responses into actionable queries to retrieve external evidence, while optimizing history reasoning traces to mitigate long-context degradation. MA-RAG extends the *self-consistency* principle by leveraging the lack of consistency as a proactive signal for multi-round agentic reasoning and retrieval, and mirrors a *boosting* mechanism that iteratively minimizes the residual error toward a stable, high-fidelity medical **consensus**. Extensive evaluations across 7 medical Q&A benchmarks show that MA-RAG consistently surpasses competitive inference-time scaling and RAG baselines, delivering **substantial +6.8 points** on average accuracy over the backbone model. Our code is available at [this url](https://github.com/NJU-RL/MA-RAG).

Paperid: 232, https://arxiv.org/pdf/2603.03290.pdf GitHub

Abstract:
Long-horizon LLM agents require memory systems that remain accurate under fixed context budgets. However, existing systems struggle with two persistent challenges in long-term dialogue: (i) \textbf{disconnected evidence}, where multi-hop answers require linking facts distributed across time, and (ii) \textbf{state updates}, where evolving information (e.g., schedule changes) creates conflicts with older static logs. We propose AriadneMem, a structured memory system that addresses these failure modes via a decoupled two-phase pipeline. In the \textbf{offline construction phase}, AriadneMem employs \emph{entropy-aware gating} to filter noise and low-information message before LLM extraction and applies \emph{conflict-aware coarsening} to merge static duplicates while preserving state transitions as temporal edges. In the \textbf{online reasoning phase}, rather than relying on expensive iterative planning, AriadneMem executes \emph{algorithmic bridge discovery} to reconstruct missing logical paths between retrieved facts, followed by \emph{single-call topology-aware synthesis}. On LoCoMo experiments with GPT-4o, AriadneMem improves \textbf{Multi-Hop F1 by 15.2\%} and \textbf{Average F1 by 9.0\%} over strong baselines. Crucially, by offloading reasoning to the graph layer, AriadneMem reduces \textbf{total runtime by 77.8\%} using only \textbf{497} context tokens. The code is available at https://github.com/LLM-VLM-GSL/AriadneMem.

Paperid: 233, https://arxiv.org/pdf/2603.03283.pdf GitHub

Abstract:
We dream of a future where point clouds from all domains can come together to shape a single model that benefits them all. Toward this goal, we present Utonia, a first step toward training a single self-supervised point transformer encoder across diverse domains, spanning remote sensing, outdoor LiDAR, indoor RGB-D sequences, object-centric CAD models, and point clouds lifted from RGB-only videos. Despite their distinct sensing geometries, densities, and priors, Utonia learns a consistent representation space that transfers across domains. This unification improves perception capability while revealing intriguing emergent behaviors that arise only when domains are trained jointly. Beyond perception, we observe that Utonia representations can also benefit embodied and multimodal reasoning: conditioning vision-language-action policies on Utonia features improves robotic manipulation, and integrating them into vision-language models yields gains on spatial reasoning. We hope Utonia can serve as a step toward foundation models for sparse 3D data, and support downstream applications in AR/VR, robotics, and autonomous driving.

Paperid: 234, https://arxiv.org/pdf/2603.03281.pdf GitHub

Abstract:
Classifier-Free Guidance (CFG) has emerged as a central approach for enhancing semantic alignment in flow-based diffusion models. In this paper, we explore a unified framework called CFG-Ctrl, which reinterprets CFG as a control applied to the first-order continuous-time generative flow, using the conditional-unconditional discrepancy as an error signal to adjust the velocity field. From this perspective, we summarize vanilla CFG as a proportional controller (P-control) with fixed gain, and typical follow-up variants develop extended control-law designs derived from it. However, existing methods mainly rely on linear control, inherently leading to instability, overshooting, and degraded semantic fidelity especially on large guidance scales. To address this, we introduce Sliding Mode Control CFG (SMC-CFG), which enforces the generative flow toward a rapidly convergent sliding manifold. Specifically, we define an exponential sliding mode surface over the semantic prediction error and introduce a switching control term to establish nonlinear feedback-guided correction. Moreover, we provide a Lyapunov stability analysis to theoretically support finite-time convergence. Experiments across text-to-image generation models including Stable Diffusion 3.5, Flux, and Qwen-Image demonstrate that SMC-CFG outperforms standard CFG in semantic alignment and enhances robustness across a wide range of guidance scales. Project Page: https://hanyang-21.github.io/CFG-Ctrl

Paperid: 235, https://arxiv.org/pdf/2603.03280.pdf GitHub

Abstract:
Many essential manipulation tasks - such as food preparation, surgery, and craftsmanship - remain intractable for autonomous robots. These tasks are characterized not only by contact-rich, force-sensitive dynamics, but also by their "implicit" success criteria: unlike pick-and-place, task quality in these domains is continuous and subjective (e.g. how well a potato is peeled), making quantitative evaluation and reward engineering difficult. We present a learning framework for such tasks, using peeling with a knife as a representative example. Our approach follows a two-stage pipeline: first, we learn a robust initial policy via force-aware data collection and imitation learning, enabling generalization across object variations; second, we refine the policy through preference-based finetuning using a learned reward model that combines quantitative task metrics with qualitative human feedback, aligning policy behavior with human notions of task quality. Using only 50-200 peeling trajectories, our system achieves over 90% average success rates on challenging produce including cucumbers, apples, and potatoes, with performance improving by up to 40% through preference-based finetuning. Remarkably, policies trained on a single produce category exhibit strong zero-shot generalization to unseen in-category instances and to out-of-distribution produce from different categories while maintaining over 90% success rates.

Paperid: 236, https://arxiv.org/pdf/2603.03275.pdf GitHub

Abstract:
Human mobility trajectories are widely studied in public health and social science, where different demographic groups exhibit significantly different mobility patterns. However, existing trajectory generation models rarely capture this heterogeneity because most trajectory datasets lack demographic labels. To address this gap in data, we propose ATLAS, a weakly supervised approach for demographic-conditioned trajectory generation using only (i) individual trajectories without demographic labels, (ii) region-level aggregated mobility features, and (iii) region-level demographic compositions from census data. ATLAS trains a trajectory generator and fine-tunes it so that simulated mobility matches observed regional aggregates while conditioning on demographics. Experiments on real trajectory data with demographic labels show that ATLAS substantially improves demographic realism over baselines (JSD $\downarrow$ 12%--69%) and closes much of the gap to strongly supervised training. We further develop theoretical analyses for when and why ATLAS works, identifying key factors including demographic diversity across regions and the informativeness of the aggregate feature, paired with experiments demonstrating the practical implications of our theory. We release our code at https://github.com/schang-lab/ATLAS.

Paperid: 237, https://arxiv.org/pdf/2603.03265.pdf GitHub

Abstract:
We present DuoMo, a generative method that recovers human motion in world-space coordinates from unconstrained videos with noisy or incomplete observations. Reconstructing such motion requires solving a fundamental trade-off: generalizing from diverse and noisy video inputs while maintaining global motion consistency. Our approach addresses this problem by factorizing motion learning into two diffusion models. The camera-space model first estimates motion from videos in camera coordinates. The world-space model then lifts this initial estimate into world coordinates and refines it to be globally consistent. Together, the two models can reconstruct motion across diverse scenes and trajectories, even from highly noisy or incomplete observations. Moreover, our formulation is general, generating the motion of mesh vertices directly and bypassing parametric models. DuoMo achieves state-of-the-art performance. On EMDB, our method obtains a 16% reduction in world-space reconstruction error while maintaining low foot skating. On RICH, it obtains a 30% reduction in world-space error. Project page: https://yufu-wang.github.io/duomo/

Paperid: 238, https://arxiv.org/pdf/2603.03203.pdf GitHub

Abstract:
CDD, or Contamination Detection via output Distribution, identifies data contamination by measuring the peakedness of a model's sampled outputs. We study the conditions under which this approach succeeds and fails on small language models ranging from 70M to 410M parameters. Using controlled contamination experiments on GSM8K, HumanEval, and MATH, we find that CDD's effectiveness depends critically on whether fine-tuning produces verbatim memorization. With low-rank adaptation, models can learn from contaminated data without memorizing it, and CDD performs at chance level even when the data is verifiably contaminated. Only when fine-tuning capacity is sufficient to induce memorization does CDD recover strong detection accuracy. Our results characterize a memorization threshold that governs detectability and highlight a practical consideration: parameter-efficient fine-tuning can produce contamination that output-distribution methods do not detect. Our code is available at https://github.com/Sela-Omer/Contamination-Detection-Small-LM

Paperid: 239, https://arxiv.org/pdf/2603.03202.pdf GitHub

Abstract:
As large language models (LLMs) advance their mathematical capabilities toward the IMO level, the scarcity of challenging, high-quality problems for training and evaluation has become a significant bottleneck. Simultaneously, recent code agents have demonstrated sophisticated skills in agentic coding and reasoning, suggesting that code execution can serve as a scalable environment for mathematical experimentation. In this paper, we investigate the potential of code agents to autonomously evolve existing math problems into more complex variations. We introduce a multi-agent framework designed to perform problem evolution while validating the solvability and increased difficulty of the generated problems. Our experiments demonstrate that, given sufficient test-time exploration, code agents can synthesize new, solvable problems that are structurally distinct from and more challenging than the originals. This work provides empirical evidence that code-driven agents can serve as a viable mechanism for synthesizing high-difficulty mathematical reasoning problems within scalable computational environments. Our data is available at https://github.com/TarferSoul/Code2Math.

Paperid: 240, https://arxiv.org/pdf/2603.03198.pdf GitHub

Abstract:
Universal embodied intelligence demands robust generalization across heterogeneous embodiments, such as autonomous driving, robotics, and unmanned aerial vehicles (UAVs). However, existing embodied brain in training a unified model over diverse embodiments frequently triggers long-tail data, gradient interference, and catastrophic forgetting, making it notoriously difficult to balance universal generalization with domain-specific proficiency. In this report, we introduce ACE-Brain-0, a generalist foundation brain that unifies spatial reasoning, autonomous driving, and embodied manipulation within a single multimodal large language model~(MLLM). Our key insight is that spatial intelligence serves as a universal scaffold across diverse physical embodiments: although vehicles, robots, and UAVs differ drastically in morphology, they share a common need for modeling 3D mental space, making spatial cognition a natural, domain-agnostic foundation for cross-embodiment transfer. Building on this insight, we propose the Scaffold-Specialize-Reconcile~(SSR) paradigm, which first establishes a shared spatial foundation, then cultivates domain-specialized experts, and finally harmonizes them through data-free model merging. Furthermore, we adopt Group Relative Policy Optimization~(GRPO) to strengthen the model's comprehensive capability. Extensive experiments demonstrate that ACE-Brain-0 achieves competitive and even state-of-the-art performance across 24 spatial and embodiment-related benchmarks.

Paperid: 241, https://arxiv.org/pdf/2603.03197.pdf GitHub

Abstract:
Classifying fine-grained visual concepts under open-world settings, i.e., without a predefined label set, demands models to be both accurate and specific. Recent reasoning Large Multimodal Models (LMMs) exhibit strong visual understanding capability but tend to produce overly generic predictions when performing fine-grained image classification. Our preliminary analysis reveals that models do possess the intrinsic fine-grained domain knowledge. However, promoting more specific predictions (specificity) without compromising correct ones (correctness) remains a non-trivial and understudied challenge. In this work, we investigate how to steer reasoning LMMs toward predictions that are both correct and specific. We propose a novel specificity-aware reinforcement learning framework, SpeciaRL, to fine-tune reasoning LMMs on fine-grained image classification under the open-world setting. SpeciaRL introduces a dynamic, verifier-based reward signal anchored to the best predictions within online rollouts, promoting specificity while respecting the model's capabilities to prevent incorrect predictions. Our out-of-domain experiments show that SpeciaRL delivers the best trade-off between correctness and specificity across extensive fine-grained benchmarks, surpassing existing methods and advancing open-world fine-grained image classification. Code and model are publicly available at https://github.com/s-angheben/SpeciaRL.

Paperid: 242, https://arxiv.org/pdf/2603.03195.pdf GitHub GitHub

Abstract:
Vision-Language-Action (VLA) models are a promising path toward embodied intelligence, yet they often overlook the predictive and temporal-causal structure underlying visual dynamics. World-model VLAs address this by predicting future frames, but waste capacity reconstructing redundant backgrounds. Latent-action VLAs encode frame-to-frame transitions compactly, but lack temporally continuous dynamic modeling and world knowledge. To overcome these limitations, we introduce CoWVLA (Chain-of-World VLA), a new "Chain of World" paradigm that unifies world-model temporal reasoning with a disentangled latent motion representation. First, a pretrained video VAE serves as a latent motion extractor, explicitly factorizing video segments into structure and motion latents. Then, during pre-training, the VLA learns from an instruction and an initial frame to infer a continuous latent motion chain and predict the segment's terminal frame. Finally, during co-fine-tuning, this latent dynamic is aligned with discrete action prediction by jointly modeling sparse keyframes and action sequences in a unified autoregressive decoder. This design preserves the world-model benefits of temporal reasoning and world knowledge while retaining the compactness and interpretability of latent actions, enabling efficient visuomotor learning. Extensive experiments on robotic simulation benchmarks show that CoWVLA outperforms existing world-model and latent-action approaches and achieves moderate computational efficiency, highlighting its potential as a more effective VLA pretraining paradigm. The project website can be found at https://fx-hit.github.io/cowvla-io.

Paperid: 243, https://arxiv.org/pdf/2603.03194.pdf GitHub GitHub

Abstract:
Current benchmarks for code agents primarily assess narrow, repository-specific fixes, overlooking critical real-world challenges such as cross-repository reasoning, domain-specialized problem solving, dependency-driven migration, and full-repository generation. To address this gap, we introduce BeyondSWE, a comprehensive benchmark that broadens existing evaluations along two axes - resolution scope and knowledge scope - using 500 real-world instances across four distinct settings. Experimental results reveal a significant capability gap: even frontier models plateau below 45% success, and no single model performs consistently across task types. To systematically investigate the role of external knowledge, we develop SearchSWE, a framework that integrates deep search with coding abilities. Our experiments show that search augmentation yields inconsistent gains and can in some cases degrade performance, highlighting the difficulty of emulating developer-like workflows that interleave search and reasoning during coding tasks. This work offers both a realistic, challenging evaluation benchmark and a flexible framework to advance research toward more capable code agents.

Paperid: 244, https://arxiv.org/pdf/2603.03187.pdf GitHub

Abstract:
Medical image segmentation commonly relies on U-shaped encoder-decoder architectures such as U-Net, where skip connections preserve fine spatial detail by injecting high-resolution encoder features into the decoder. However, these skip pathways also propagate low-level textures, background clutter, and acquisition noise, allowing irrelevant information to bypass deeper semantic filtering -- an issue that is particularly detrimental in low-contrast clinical imaging. Although attention gates have been introduced to address this limitation, they typically produce dense sigmoid masks that softly reweight features rather than explicitly removing irrelevant activations. We propose ProSMA-UNet (Proximal-Sparse Multi-Scale Attention U-Net), which reformulates skip gating as a decoder-conditioned sparse feature selection problem. ProSMA constructs a multi-scale compatibility field using lightweight depthwise dilated convolutions to capture relevance across local and contextual scales, then enforces explicit sparsity via an $\ell_1$ proximal operator with learnable per-channel thresholds, yielding a closed-form soft-thresholding gate that can remove noisy responses. To further suppress semantically irrelevant channels, ProSMA incorporates decoder-conditioned channel gating driven by global decoder context. Extensive experiments on challenging 2D and 3D benchmarks demonstrate state-of-the-art performance, with particularly large gains ($\approx20$\%) on difficult 3D segmentation tasks. Project page: https://math-ml-x.github.io/ProSMA-UNet/

Paperid: 245, https://arxiv.org/pdf/2603.03177.pdf GitHub

Abstract:
The integration of symbolic computing with neural networks has intrigued researchers since the first theorizations of Artificial intelligence (AI). The ability of Neuro-Symbolic (NeSy) methods to infer or exploit behavioral schema has been widely considered as one of the possible proxies for human-level intelligence. However, the limited semantic generalizability and the challenges in declining complex domains with pre-defined patterns and rules hinder their practical implementation in real-world scenarios. The unprecedented results achieved by connectionist systems since the last AI breakthrough in 2017 have raised questions about the competitiveness of NeSy solutions, with particular emphasis on the Natural Language Processing and Computer Vision fields. This survey examines task-specific advancements in the NeSy domain to explore how incorporating symbolic systems can enhance explainability and reasoning capabilities. Our findings are meant to serve as a resource for researchers exploring explainable NeSy methodologies for real-life tasks and applications. Reproducibility details and in-depth comments on each surveyed research work are made available at https://github.com/disi-unibo-nlp/task-oriented-neuro-symbolic.git.

Paperid: 246, https://arxiv.org/pdf/2603.03158.pdf GitHub

Abstract:
Bengali remains a low-resource language in speech technology, especially for complex tasks like long-form transcription and speaker diarization. This paper presents a multistage approach developed for the "DL Sprint 4.0 - Bengali Long-Form Speech Recognition" and "DL Sprint 4.0 - Bengali Speaker Diarization" competitions on Kaggle, addressing the challenge of "who spoke when/what" in hour-long recordings. We implemented Whisper Medium fine-tuned on Bengali data (bengaliAI/tugstugi bengaliai-asr whisper-medium) for transcription and integrated pyannote/speaker-diarization-community-1 with our custom-trained segmentation model to handle diverse and noisy acoustic environments. Using a two-pass method with hyperparameter tuning, we achieved a DER of 0.27 on the private leaderboard and 0.19 on the public leaderboard. For transcription, chunking, background noise cleaning, and algorithmic post-processing yielded a WER of 0.38 on the private leaderboard. These results show that targeted tuning and strategic data utilization can significantly improve AI inclusivity for South Asian languages. All relevant code is available at: https://github.com/Short-Potatoes/Bengali-long-form-transcription-and-diarization.git Index Terms: Bengali speech recognition, speaker diarization, Whisper, ASR, low-resource languages, pyannote, voice activity detection

Paperid: 247, https://arxiv.org/pdf/2603.03126.pdf GitHub

Abstract:
Scholarly data are largely fragmented across siloed databases with divergent metadata and missing linkages among them. We present the Science Data Lake, a locally-deployable infrastructure built on DuckDB and simple Parquet files that unifies eight open sources - Semantic Scholar, OpenAlex, SciSciNet, Papers with Code, Retraction Watch, Reliance on Science, a preprint-to-published mapping, and Crossref - via DOI normalization while preserving source-level schemas. The resource comprises approximately 960GB of Parquet files spanning ~293 million uniquely identifiable papers across ~22 schemas and ~153 SQL views. An embedding-based ontology alignment using BGE-large sentence embeddings maps 4,516 OpenAlex topics to 13 scientific ontologies (~1.3 million terms), yielding 16,150 mappings covering 99.8% of topics ($\geq 0.65$ threshold) with $F1 = 0.77$ at the recommended $\geq 0.85$ operating point, outperforming TF-IDF, BM25, and Jaro-Winkler baselines on a 300-pair gold-standard evaluation. We validate through 10 automated checks, cross-source citation agreement analysis (pairwise Pearson $r = 0.76$ - $0.87$), and stratified manual annotation. Four vignettes demonstrate cross-source analyses infeasible with any single database. The resource is open source, deployable on a single drive or queryable remotely via HuggingFace, and includes structured documentation suitable for large language model (LLM) based research agents.

Paperid: 248, https://arxiv.org/pdf/2603.03101.pdf GitHub

Abstract:
The CLIP model's outstanding generalization has driven recent success in Zero-Shot Anomaly Detection (ZSAD) for detecting anomalies in unseen categories. The core challenge in ZSAD is to specialize the model for anomaly detection tasks while preserving CLIP's powerful generalization capability. Existing approaches attempting to solve this challenge share the fundamental limitation of a patch-agnostic design that processes all patches monolithically without regard for their unique characteristics. To address this limitation, we propose MoECLIP, a Mixture-of-Experts (MoE) architecture for the ZSAD task, which achieves patch-level adaptation by dynamically routing each image patch to a specialized Low-Rank Adaptation (LoRA) expert based on its unique characteristics. Furthermore, to prevent functional redundancy among the LoRA experts, we introduce (1) Frozen Orthogonal Feature Separation (FOFS), which orthogonally separates the input feature space to force experts to focus on distinct information, and (2) a simplex equiangular tight frame (ETF) loss to regulate the expert outputs to form maximally equiangular representations. Comprehensive experimental results across 14 benchmark datasets spanning industrial and medical domains demonstrate that MoECLIP outperforms existing state-of-the-art methods. The code is available at https://github.com/CoCoRessa/MoECLIP.

Paperid: 249, https://arxiv.org/pdf/2603.03065.pdf GitHub

Abstract:
Dense retrieval services increasingly underpin semantic search, recommendation, and retrieval-augmented generation, yet clients typically receive only a top-$k$ list with no auditable evidence of how it was produced. We present V3DB, a verifiable, versioned vector-search service that enables audit-on-demand correctness checks for approximate nearest-neighbour (ANN) retrieval executed by a potentially untrusted service provider. V3DB commits to each corpus snapshot and standardises an IVF-PQ search pipeline into a fixed-shape, five-step query semantics. Given a public snapshot commitment and a query embedding, the service returns the top-$k$ payloads and, when challenged, produces a succinct zero-knowledge proof that the output is exactly the result of executing the published semantics on the committed snapshot -- without revealing the embedding corpus or private index contents. To make proving practical, V3DB avoids costly in-circuit sorting and random access by combining multiset equality/inclusion checks with lightweight boundary conditions. Our prototype implementation based on Plonky2 achieves up to $22\times$ faster proving and up to $40\%$ lower peak memory consumption than the circuit-only baseline, with millisecond-level verification time. Github Repo at https://github.com/TabibitoQZP/zk-IVF-PQ.

Paperid: 250, https://arxiv.org/pdf/2603.03054.pdf GitHub

Abstract:
Large language models are increasingly used for patient-facing medical assistance and clinical decision support, but adapting them to clinical dialogue often requires supervision derived from doctor-patient conversations that may contain sensitive information. Conventional supervised fine-tuning and reinforcement learning from human feedback (RLHF) can amplify memorization risks, enabling empirical membership inference and extraction of rare training-set content. We present PrivMedChat, an end-to-end framework for differentially private RLHF (DP-RLHF) for medical dialogue. Our design enforces differential privacy at every training stage that directly accesses dialogue-derived supervision: (i) Differential Private Stochastic Gradient Descent (DP-SGD) for medical SFT and (ii) DP-SGD for reward model learning from preference pairs. To limit additional privacy expenditure during alignment, we apply DP-SGD to the PPO actor and critic when operating on dialogue-derived prompts, while the reward model remains fixed after DP training. We also introduce an annotation-free preference construction strategy that pairs physician responses with filtered non-expert generations to produce scalable preference data without clinician labeling. Experiments on medical dialogue benchmarks show that PrivMedChat at $\varepsilon=7$ achieves the highest ROUGE-L of 0.156 among all DP models, reduces clinical hallucinations to 1.4% and harmful advice to 0.4%, and obtains the highest overall score of 2.86 in a 3-model LLM-jury evaluation, while producing membership-inference signals that are near chance (AUC 0.510-0.555). We open-source our code at https://github.com/sudip-bhujel/privmedchat.

Paperid: 251, https://arxiv.org/pdf/2603.03031.pdf GitHub

Abstract:
Large Language Models (LLMs) have achieved strong complex reasoning capabilities through Chain-of-Thought (CoT) reasoning. However, their reasoning patterns remain too complicated to analyze. While Sparse Autoencoders (SAEs) have emerged as a powerful tool for interpretability, existing approaches predominantly operate at the token level, creating a granularity mismatch when capturing more critical step-level information, such as reasoning direction and semantic transitions. In this work, we propose step-level sparse autoencoder (SSAE), which serves as an analytical tool to disentangle different aspects of LLMs' reasoning steps into sparse features. Specifically, by precisely controlling the sparsity of a step feature conditioned on its context, we form an information bottleneck in step reconstruction, which splits incremental information from background information and disentangles it into several sparsely activated dimensions. Experiments on multiple base models and reasoning tasks show the effectiveness of the extracted features. By linear probing, we can easily predict surface-level information, such as generation length and first token distribution, as well as more complicated properties, such as the correctness and logicality of the step. These observations indicate that LLMs should already at least partly know about these properties during generation, which provides the foundation for the self-verification ability of LLMs. The code is available at https://github.com/Miaow-Lab/SSAE

Paperid: 252, https://arxiv.org/pdf/2603.03026.pdf GitHub

Abstract:
Joint estimation of surface normals and depth is essential for holistic 3D scene understanding, yet high-resolution prediction remains difficult due to the trade-off between preserving fine local detail and maintaining global consistency. To address this challenge, we propose the Ultra Resolution Geometry Transformer (URGT), which adapts the Visual Geometry Grounded Transformer (VGGT) into a unified multi-patch transformer for monocular high-resolution depth--normal estimation. A single high-resolution image is partitioned into patches that are augmented with coarse depth and normal priors from pre-trained models, and jointly processed in a single forward pass to predict refined geometric outputs. Global coherence is enforced through cross-patch attention, which enables long-range geometric reasoning and seamless propagation of information across patches within a shared backbone. To further enhance spatial robustness, we introduce a GridMix patch sampling strategy that probabilistically samples grid configurations during training, improving inter-patch consistency and generalization. Our method achieves state-of-the-art results on UnrealStereo4K, jointly improving depth and normal estimation, reducing AbsRel from 0.0582 to 0.0291, RMSE from 2.17 to 1.31, and lowering mean angular error from 23.36 degrees to 18.51 degrees, while producing sharper and more stable geometry. The proposed multi-patch framework also demonstrates strong zero-shot and cross-domain generalization and scales effectively to very high resolutions, offering an efficient and extensible solution for high-quality geometry refinement.

Paperid: 253, https://arxiv.org/pdf/2603.03017.pdf GitHub

Abstract:
In this work, we address the need for efficient and formally stable Recurrent Neural Networks (RNNs) in environments with limited computational resources by analyzing the stability of the Minimal Gated Unit (MGU) network, a lightweight alternative to common gated RNNs used in system identification. We derive sufficient parametric conditions for the MGU network's input-to-state stability and incremental input-to-state stability properties. These conditions enable a-posteriori validation of model stability and form the basis for novel stability-promoting training methodologies, including a warm-start of the network's parameters and a projected gradient-based optimization scheme, both of which are presented in this work. Comparative evaluation, including robustness analysis and validation on synthetic and real-world data (i.e., the Silverbox benchmark), demonstrates that the minimal gated unit network successfully combines formal stability guarantees with superior parameter efficiency and faster inference times compared to other state-of-the-art recurrent neural networks, while maintaining comparable and satisfactory accuracy. Notably, the results attained on the Silverbox benchmark illustrate that the stable MGU network effectively captures the system dynamics, whereas other stable RNNs fail to converge to a reliable model.

Paperid: 254, https://arxiv.org/pdf/2603.02974.pdf GitHub

Abstract:
DINO models provide rich patch-level representations that have recently enabled strong performance in unsupervised anomaly detection (UAD). Most existing methods extract patch embeddings from ``normal'' images and model them independently, ignoring spatial and neighborhood relationships between patches. This implicitly assumes that self-attention and positional encodings sufficiently encode contextual information within each patch embedding. In addition, the normative distribution is often modeled as memory banks or prototype-based representations, which require storing large numbers of features and performing costly comparisons at inference time, leading to substantial memory and computational overhead. In this work, we address both limitations by proposing a simple and efficient framework that explicitly models spatial and contextual dependencies between patch embeddings using a 2D autoregressive (AR) model. Instead of storing embeddings or clustering prototypes, our approach learns a compact parametric model of the normative distribution via an AR convolutional neural network (CNN). At test time, anomaly detection reduces to a single forward pass through the network and enables fast and memory-efficient inference. We evaluate our method on the BMAD benchmark, which comprises three medical imaging datasets, and compare it against existing work including recent DINO-based methods. Experimental results demonstrate that explicitly modeling spatial dependencies achieves competitive anomaly detection performance while substantially reducing inference time and memory requirements. Code is available at the project page: https://eerdil.github.io/spatial-ar-dinov3-uad/.

Paperid: 255, https://arxiv.org/pdf/2603.02972.pdf GitHub

Abstract:
Vision-Language Navigation (VLN) presents a unique challenge for Large Vision-Language Models (VLMs) due to their inherent architectural mismatch: VLMs are primarily pretrained on static, disembodied vision-language tasks, which fundamentally clash with the dynamic, embodied, and spatially-structured nature of navigation. Existing large-model-based methods often resort to converting rich visual and spatial information into text, forcing models to implicitly infer complex visual-topological relationships or limiting their global action capabilities. To bridge this gap, we propose TagaVLM (Topology-Aware Global Action reasoning), an end-to-end framework that explicitly injects topological structures into the VLM backbone. To introduce topological edge information, Spatial Topology Aware Residual Attention (STAR-Att) directly integrates it into the VLM's self-attention mechanism, enabling intrinsic spatial reasoning while preserving pretrained knowledge. To enhance topological node information, an Interleaved Navigation Prompt strengthens node-level visual-text alignment. Finally, with the embedded topological graph, the model is capable of global action reasoning, allowing for robust path correction. On the R2R benchmark, TagaVLM achieves state-of-the-art performance among large-model-based methods, with a Success Rate (SR) of 51.09% and SPL of 47.18 in unseen environments, outperforming prior work by 3.39% in SR and 9.08 in SPL. This demonstrates that, for embodied spatial reasoning, targeted enhancements on smaller open-source VLMs can be more effective than brute-force model scaling. The code will be released upon publication.Project page: https://apex-bjut.github.io/Taga-VLM

Paperid: 256, https://arxiv.org/pdf/2603.02959.pdf GitHub

Abstract:
Vision-language models (VLMs) pre-trained on large, heterogeneous data sources are becoming increasingly popular, providing rich multi-modal embeddings that enable efficient transfer to new tasks. A particularly relevant application is few-shot adaptation, where only a handful of annotated examples are available to adapt the model through multi-modal linear probes. In medical imaging, specialized VLMs have shown promising performance in zero- and few-shot image classification, which is valuable for mitigating the high cost of expert annotations. However, challenges remain in extremely low-shot regimes: the inherent class imbalances in medical tasks often lead to underrepresented categories, penalizing overall model performance. To address this limitation, we propose leveraging unlabeled data by introducing an efficient semi-supervised solver that propagates text-informed pseudo-labels during few-shot adaptation. The proposed method enables lower-budget annotation pipelines for adapting VLMs, reducing labeling effort by >50% in low-shot regimes.

Paperid: 257, https://arxiv.org/pdf/2603.02924.pdf GitHub

Abstract:
Despite the growing interest in open-vocabulary object detection in recent years, most existing methods rely heavily on manually curated fine-grained training datasets as well as resource-intensive layer-wise cross-modal feature extraction. In this paper, we propose HDINO, a concise yet efficient open-vocabulary object detector that eliminates the dependence on these components. Specifically, we propose a two-stage training strategy built upon the transformer-based DINO model. In the first stage, noisy samples are treated as additional positive object instances to construct a One-to-Many Semantic Alignment Mechanism(O2M) between the visual and textual modalities, thereby facilitating semantic alignment. A Difficulty Weighted Classification Loss (DWCL) is also designed based on initial detection difficulty to mine hard examples and further improve model performance. In the second stage, a lightweight feature fusion module is applied to the aligned representations to enhance sensitivity to linguistic semantics. Under the Swin Transformer-T setting, HDINO-T achieves \textbf{49.2} mAP on COCO using 2.2M training images from two publicly available detection datasets, without any manual data curation and the use of grounding data, surpassing Grounding DINO-T and T-Rex2 by \textbf{0.8} mAP and \textbf{2.8} mAP, respectively, which are trained on 5.4M and 6.5M images. After fine-tuning on COCO, HDINO-T and HDINO-L further achieve \textbf{56.4} mAP and \textbf{59.2} mAP, highlighting the effectiveness and scalability of our approach. Code and models are available at https://github.com/HaoZ416/HDINO.

Paperid: 258, https://arxiv.org/pdf/2603.02910.pdf GitHub GitHub

Abstract:
Articulated objects are ubiquitous in daily life. Our goal is to achieve a high-quality reconstruction, segmentation of independent moving parts, and analysis of articulation. Recent methods analyse two different articulation states and perform per-point part segmentation, optimising per-part articulation using cross-state correspondences, given a priori knowledge of the number of parts. Such assumptions greatly limit their applications and performance. Their robustness is reduced when objects cannot be clearly visible in both states. To address these issues, in this paper, we present a new framework, Articulation in Motion (AiM). We infer part-level decomposition, articulation kinematics, and reconstruct an interactive 3D digital replica from a user-object interaction video and a start-state scan. We propose a dual-Gaussian scene representation that is learned from an initial 3DGS scan of the object and a video that shows the movement of separate parts. It uses motion cues to segment the object into parts and assign articulation joints. Subsequently, a robust, sequential RANSAC is employed to achieve part mobility analysis without any part-level structural priors, which clusters moving primitives into rigid parts and estimates kinematics while automatically determining the number of parts. The proposed approach separates the object into parts, each represented as a 3D Gaussian set, enabling high-quality rendering. Our approach yields higher quality part segmentation than previous methods, without prior knowledge. Extensive experimental analysis on both simple and complex objects validates the effectiveness and strong generalisation ability of our approach. Project page: https://haoai-1997.github.io/AiM/.

Paperid: 259, https://arxiv.org/pdf/2603.02908.pdf GitHub

Abstract:
In recent years, pre-trained large language models have achieved remarkable success across diverse tasks. Besides the pivotal role of self-supervised pre-training, their effectiveness in downstream applications also depends critically on the post-training process, which adapts models to task-specific data and objectives. However, this process inevitably introduces model shifts that can influence performance in different domains, and how such shifts transfer remains poorly understood. To open up the black box, we propose the SAE-based Transferability Score (STS), a new metric that leverages sparse autoencoders (SAEs) to forecast post-training transferability. Taking supervised fine-tuning as an example, STS identifies shifted dimensions in SAE representations and calculates their correlations with downstream domains, enabling reliable estimation of transferability \textit{before} fine-tuning. Extensive experiments across multiple models and domains show that STS accurately predicts the transferability of supervised fine-tuning, achieving Pearson correlation coefficients above 0.7 with actual performance changes. Beyond this, we take an initial step toward extending STS to reinforcement learning. We believe that STS can serve as an {\color{black} interpretable} tool for guiding post-training strategies in LLMs. Code is available at https://github.com/PKU-ML/STS.

Paperid: 260, https://arxiv.org/pdf/2603.02882.pdf GitHub

Abstract:
Artificial Intelligence Generated Content (AIGC), particularly video generation with diffusion models, has been advanced rapidly. Invisible watermarking is a key technology for protecting AI-generated videos and tracing harmful content, and thus plays a crucial role in AI safety. Beyond post-processing watermarks which inevitably degrade video quality, recent studies have proposed distortion-free in-generation watermarking for video diffusion models. However, existing in-generation approaches are non-blind: they require maintaining all the message-key pairs and performing template-based matching during extraction, which incurs prohibitive computational costs at scale. Moreover, when applied to modern video diffusion models with causal 3D Variational Autoencoders (VAEs), their robustness against temporal disturbance becomes extremely weak. To overcome these challenges, we propose SIGMark, a Scalable In-Generation watermarking framework with blind extraction for video diffusion. To achieve blind-extraction, we propose to generate watermarked initial noise using a Global set of Frame-wise PseudoRandom Coding keys (GF-PRC), reducing the cost of storing large-scale information while preserving noise distribution and diversity for distortion-free watermarking. To enhance robustness, we further design a Segment Group-Ordering module (SGO) tailored to causal 3D VAEs, ensuring robust watermark inversion during extraction under temporal disturbance. Comprehensive experiments on modern diffusion models show that SIGMark achieves very high bit-accuracy during extraction under both temporal and spatial disturbances with minimal overhead, demonstrating its scalability and robustness. Our project is available at https://jeremyzhao1998.github.io/SIGMark-release/.

Paperid: 261, https://arxiv.org/pdf/2603.02872.pdf GitHub

Abstract:
Large Vision Language Models (LVLMs) exhibit strong Chain-of-Thought (CoT) capabilities, yet most existing paradigms assume full-video availability before inference, a batch-style process misaligned with real-world video streams where information arrives sequentially. Motivated by the streaming nature of video data, we investigate two streaming reasoning paradigms for LVLMs. The first, an interleaved paradigm, alternates between receiving frames and producing partial reasoning but remains constrained by strictly ordered cache updates. To better match streaming inputs, we propose \textbf{Think-as-You-See (TaYS)}, a unified framework enabling true concurrent reasoning. TaYS integrates parallelized CoT generation, stream-constrained training, and stream-parallel inference. It further employs temporally aligned reasoning units, streaming attention masks and positional encodings, and a dual KV-cache that decouples visual encoding from textual reasoning. We evaluate all paradigms on the Qwen2.5-VL family across representative video CoT tasks, including event dynamics analysis, causal reasoning, and thematic understanding. Experiments show that TaYS consistently outperforms both batch and interleaved baselines, improving reasoning performance while substantially reducing time-to-first-token (TTFT) and overall reasoning delay. These results demonstrate the effectiveness of data-aligned streaming reasoning in enabling efficient and responsive video understanding for LVLMs. We release our code at https://github.com/EIT-NLP/StreamingLLM/tree/main/TaYS

Paperid: 262, https://arxiv.org/pdf/2603.02829.pdf GitHub

Abstract:
Recent text-to-image (T2I) diffusion and flow-matching models can produce highly realistic images from natural language prompts. In practical scenarios, T2I systems are often run in a ``generate--then--select'' mode: many seeds are sampled and only a few images are kept for use. However, this pipeline is highly resource-intensive since each candidate requires tens to hundreds of denoising steps, and evaluation metrics such as CLIPScore and ImageReward are post-hoc. In this work, we address this inefficiency by introducing Probe-Select, a plug-in module that enables efficient evaluation of image quality within the generation process. We observe that certain intermediate denoiser activations, even at early timesteps, encode a stable coarse structure, object layout and spatial arrangement--that strongly correlates with final image fidelity. Probe-Select exploits this property by predicting final quality scores directly from early activations, allowing unpromising seeds to be terminated early. Across diffusion and flow-matching backbones, our experiments show that early evaluation at only 20\% of the trajectory accurately ranks candidate seeds and enables selective continuation. This strategy reduces sampling cost by over 60\% while improving the quality of the retained images, demonstrating that early structural signals can effectively guide selective generation without altering the underlying generative model. Code is available at https://github.com/Guhuary/ProbeSelect.

Paperid: 263, https://arxiv.org/pdf/2603.02787.pdf GitHub

Abstract:
The rise of Large Language Model-based Automated Algorithm Design (LLM-AAD) has transformed algorithm development by autonomously generating code implementations of expert-level algorithms. Unlike traditional expert-driven algorithm development, in the LLM-AAD paradigm, the main design principle behind an algorithm is often implicitly embedded in the generated code. Therefore, assessing algorithmic similarity directly from code, distinguishing genuine algorithmic innovation from mere syntactic variation, becomes essential. While various code similarity metrics exist, they fail to capture algorithmic similarity, as they focus on surface-level syntax or output equivalence rather than the underlying algorithmic logic. We propose BehaveSim, a novel method to measure algorithmic similarity through the lens of problem-solving behavior as a sequence of intermediate solutions produced during execution, dubbed as problem-solving trajectories (PSTrajs). By quantifying the alignment between PSTrajs using dynamic time warping (DTW), BehaveSim distinguishes algorithms with divergent logic despite syntactic or output-level similarities. We demonstrate its utility in two key applications: (i) Enhancing LLM-AAD: Integrating BehaveSim into existing LLM-AAD frameworks (e.g., FunSearch, EoH) promotes behavioral diversity, significantly improving performance on three AAD tasks. (ii) Algorithm analysis: BehaveSim clusters generated algorithms by behavior, enabling systematic analysis of problem-solving strategies--a crucial tool for the growing ecosystem of AI-generated algorithms. Data and code of this work are open-sourced at https://github.com/RayZhhh/behavesim.

Paperid: 264, https://arxiv.org/pdf/2603.02754.pdf GitHub

Abstract:
Multimodal large language models (MLLMs) suffer from pronounced hallucinations in remote sensing visual question-answering (RS-VQA), primarily caused by visual grounding failures in large-scale scenes or misinterpretation of fine-grained small targets. To systematically analyze these issues, we introduce RSHBench, a protocol-based benchmark for fine-grained diagnosis of factual and logical hallucinations. To mitigate grounding-induced factual hallucinations, we further propose Relative Attention-Driven Actively Reasoning (RADAR), a training-free inference method that leverages intrinsic attention in MLLMs to guide progressive localization and fine-grained local reasoning at test time. Extensive experiments across diverse MLLMs demonstrate that RADAR consistently improves RS-VQA performance and reduces both factual and logical hallucinations. Code and data will be publicly available at: https://github.com/MiliLab/RADAR

Paperid: 265, https://arxiv.org/pdf/2603.02753.pdf GitHub GitHub

Abstract:
Designing novel proteins with desired characteristics remains a significant challenge due to the large sequence space and the complexity of sequence-function relationships. Efficient exploration of this space to identify sequences that meet specific design criteria is crucial for advancing therapeutics and biotechnology. Here, we present BoGA (Bayesian Optimization Genetic Algorithm), a framework that combines evolutionary search with Bayesian optimization to efficiently navigate the sequence space. By integrating a genetic algorithm as a stochastic proposal generator within a surrogate modeling loop, BoGA prioritizes candidates based on prior evaluations and surrogate model predictions, enabling data-efficient optimization. We demonstrate the utility of BoGA through benchmarking on sequence and structure design tasks, followed by its application in designing peptide binders against pneumolysin, a key virulence factor of \textit{Streptococcus pneumoniae}. BoGA accelerates the discovery of high-confidence binders, demonstrating the potential for efficient protein design across diverse objectives. The algorithm is implemented within the BoPep suite and is available under an MIT license at \href{https://github.com/ErikHartman/bopep}{GitHub}.

Paperid: 266, https://arxiv.org/pdf/2603.02730.pdf GitHub

Abstract:
Generative recommendation has recently emerged as a promising paradigm in sequential recommendation. It formulates the task as an autoregressive generation process, predicting discrete tokens of the next item conditioned on user interaction histories. Existing generative recommendation models are typically trained with token-level likelihood objectives, such as cross-entropy loss, while employing multi-step beam search during inference to generate ranked item candidates. However, this leads to a fundamental training-inference inconsistency: standard training assumes ground-truth history is always available, ignoring the fact that beam search prunes low-probability branches during inference. Consequently, the correct item may be prematurely discarded simply because its initial tokens (prefixes) have low scores. To address this issue, we propose the Adaptive Prefix-Aware Optimization (APAO) framework, which introduces prefix-level optimization losses to better align the training objective with the inference setting. Furthermore, we design an adaptive worst-prefix optimization strategy that dynamically focuses on the most vulnerable prefixes during training, thereby enhancing the model's ability to retain correct candidates under beam search constraints. We provide theoretical analyses to demonstrate the effectiveness and efficiency of our framework. Extensive experiments on multiple datasets further show that APAO consistently alleviates the training-inference inconsistency and improves performance across various generative recommendation backbones. Our codes are publicly available at https://github.com/yuyq18/APAO.

Paperid: 267, https://arxiv.org/pdf/2603.02726.pdf GitHub

Abstract:
Cross-view geo-localization (CVGL) aims to establish spatial correspondences between images captured from significantly different viewpoints and constitutes a fundamental technique for visual localization in GNSS-denied environments. Nevertheless, CVGL remains challenging due to severe geometric asymmetry, texture inconsistency across imaging domains, and the progressive degradation of discriminative local information. Existing methods predominantly rely on spatial domain feature alignment, which is inherently sensitive to large scale viewpoint variations and local disturbances. To alleviate these limitations, this paper proposes the Spatial and Frequency Domain Enhancement Network (SFDE), which leverages complementary representations from spatial and frequency domains. SFDE adopts a three branch parallel architecture to model global semantic context, local geometric structure, and statistical stability in the frequency domain, respectively, thereby characterizing consistency across domains from the perspectives of scene topology, multiscale structural patterns, and frequency invariance. The resulting complementary features are jointly optimized in a unified embedding space via progressive enhancement and coupled constraints, enabling the learning of cross-view representations with consistency across multiple granularities. Comprehensive experiments show that SFDE achieves competitive performance and in many cases even surpasses state-of-the-art methods, while maintaining a lightweight and computationally efficient design. {Our code is available at https://github.com/Mashuaishuai669/SFDE

Paperid: 268, https://arxiv.org/pdf/2603.02710.pdf GitHub

Abstract:
All-in-one image restoration is challenging because different degradation types, such as haze, blur, noise, and low-light, impose diverse requirements on restoration strategies, making it difficult for a single model to handle them effectively. In this paper, we propose a unified image restoration framework that integrates a dual-level Mixture-of-Experts (MoE) architecture with a pretrained diffusion model. The framework operates at two levels: the Inter-MoE layer adaptively combines expert groups to handle major degradation types, while the Intra-MoE layer further selects specialized sub-experts to address fine-grained variations within each type. This design enables the model to achieve coarse-grained adaptation across diverse degradation categories while performing fine-grained modulation for specific intra-class variations, ensuring both high specialization in handling complex, real-world corruptions. Extensive experiments demonstrate that the proposed method performs favorably against the state-of-the-art approaches on multiple image restoration task.

Paperid: 269, https://arxiv.org/pdf/2603.02692.pdf GitHub

Abstract:
Diffusion-based approaches have recently driven remarkable progress in real-world image super-resolution (SR). However, existing methods still struggle to simultaneously preserve fine details and ensure high-fidelity reconstruction, often resulting in suboptimal visual quality. In this paper, we propose FiDeSR, a high-fidelity and detail-preserving one-step diffusion super-resolution framework. During training, we introduce a detail-aware weighting strategy that adaptively emphasizes regions where the model exhibits higher prediction errors. During inference, low- and high-frequency adaptive enhancers further refine the reconstruction without requiring model retraining, enabling flexible enhancement control. To further improve the reconstruction accuracy, FiDeSR incorporates a residual-in-residual noise refinement, which corrects prediction errors in the diffusion noise and enhances fine detail recovery. FiDeSR achieves superior real-world SR performance compared to existing diffusion-based methods, producing outputs with both high perceptual quality and faithful content restoration. The source code will be released at: https://github.com/Ar0Kim/FiDeSR.

Paperid: 270, https://arxiv.org/pdf/2603.02640.pdf GitHub

Abstract:
Online platforms increasingly rely on opinion aggregation to allocate real-world attention and resources, yet common signals such as engagement votes or capital-weighted commitments are easy to amplify and often track visibility rather than reliability. This makes collective judgments brittle under weak truth signals, noisy or delayed feedback, early popularity surges, and strategic manipulation. We propose Credibility Governance (CG), a mechanism that reallocates influence by learning which agents and viewpoints consistently track evolving public evidence. CG maintains dynamic credibility scores for both agents and opinions, updates opinion influence via credibility-weighted endorsements, and updates agent credibility based on the long-run performance of the opinions they support, rewarding early and persistent alignment with emerging evidence while filtering short-lived noise. We evaluate CG in POLIS, a socio-physical simulation environment that models coupled belief dynamics and downstream feedback under uncertainty. Across settings with initial majority misalignment, observation noise and contamination, and misinformation shocks, CG outperforms vote-based, stake-weighted, and no-governance baselines, yielding faster recovery to the true state, reduced lock-in and path dependence, and improved robustness under adversarial pressure. Our implementation and experimental scripts are publicly available at https://github.com/Wanying-He/Credibility_Governance.

Paperid: 271, https://arxiv.org/pdf/2603.02635.pdf GitHub

Abstract:
Vision-language models remain susceptible to multimodal jailbreaks and over-refusal because safety hinges on both visual evidence and user intent, while many alignment pipelines supervise only the final response. To address this, we present SaFeR-ToolKit, which formalizes safety decision-making as a checkable protocol. Concretely, a planner specifies a persona, a Perception $\to$ Reasoning $\to$ Decision tool set, and a constrained transition graph, while a responder outputs a typed key-value tool trace before the final answer. To make the protocol reliably followed in practice, we train a single policy with a three-stage curriculum (SFT $\to$ DPO $\to$ GRPO), where GRPO directly supervises tool usage beyond answer-level feedback. Our contributions are two-fold: I. Dataset. The first tool-based safety reasoning dataset, comprising 31,654 examples (SFT 6k, DPO 18.6k, GRPO 6k) plus 1k held-out evaluation. II. Experiments. On Qwen2.5-VL, SaFeR-ToolKit significantly improves Safety/Helpfulness/Reasoning Rigor on 3B (29.39/45.04/4.98 $\to$ 84.40/71.13/78.87) and 7B (53.21/52.92/19.26 $\to$ 86.34/80.79/85.34), while preserving general capabilities (3B: 58.67 $\to$ 59.21; 7B: 66.39 $\to$ 66.81). Codes are available at https://github.com/Duebassx/SaFeR_ToolKit.

Paperid: 272, https://arxiv.org/pdf/2603.02626.pdf GitHub

Abstract:
Autonomous web navigation requires agents to perceive complex visual environments and maintain long-term context, yet current Large Language Model (LLM) based agents often struggle with spatial disorientation and navigation loops. In this paper, we propose generally applicable V-GEMS(Visual Grounding and Explicit Memory System), a robust multimodal agent architecture designed for precise and resilient web traversal. Our agent integrates visual grounding to resolve ambiguous interactive elements and introduces an explicit memory stack with state tracking. This dual mechanism allows the agent to maintain a structured map of its traversal path, enabling valid backtracking and preventing cyclical failures in deep navigation tasks. We also introduce an updatable dynamic benchmark to rigorously evaluate adaptability. Experiments show V-GEMS significantly dominates the WebWalker baseline, achieving a substantial 28.7% performance gain. Code is available at https://github.com/Vaultttttttttttt/V-GEMS.

Paperid: 273, https://arxiv.org/pdf/2603.02619.pdf GitHub GitHub

Abstract:
Single-view 3D human reconstruction has achieved remarkable progress through the adoption of multi-view diffusion models, yet the recovered 3D humans often exhibit unnatural poses. This phenomenon becomes pronounced when reconstructing 3D humans with dynamic or challenging poses, which we attribute to the limited scale of available 3D human datasets with diverse poses. To address this limitation, we introduce DrPose, Direct Reward fine-tuning algorithm on Poses, which enables post-training of a multi-view diffusion model on diverse poses without requiring expensive 3D human assets. DrPose trains a model using only human poses paired with single-view images, employing a direct reward fine-tuning to maximize PoseScore, which is our proposed differentiable reward that quantifies consistency between a generated multi-view latent image and a ground-truth human pose. This optimization is conducted on DrPose15K, a novel dataset that was constructed from an existing human motion dataset and a pose-conditioned video generative model. Constructed from abundant human pose sequence data, DrPose15K exhibits a broader pose distribution compared to existing 3D human datasets. We validate our approach through evaluation on conventional benchmark datasets, in-the-wild images, and a newly constructed benchmark, with a particular focus on assessing performance on challenging human poses. Our results demonstrate consistent qualitative and quantitative improvements across all benchmarks. Project page: https://seunguk-do.github.io/drpose.

Paperid: 274, https://arxiv.org/pdf/2603.02615.pdf GitHub

Abstract:
This project reproduces and extends the recently proposed ``Recursive Language Models'' (RLMs) framework by Zhang et al. (2026). This framework enables Large Language Models (LLMs) to process near-infinite contexts by offloading the prompt into an external REPL environment. While the original paper relies on a default recursion depth of 1 and suggests deeper recursion as a future direction, this study specifically investigates the impact of scaling the recursion depth. Using state-of-the-art open-source agentic models (DeepSeek v3.2 and Kimi K2), I evaluated pure LLM, RLM (depth=1), and RLM (depth=2) on the S-NIAH and OOLONG benchmarks. The findings reveal a compelling phenomenon: Deeper recursion causes models to ``overthink''. While depth-1 RLMs effectively boost accuracy on complex reasoning tasks, applying deeper recursion (depth=2) or using RLMs on simple retrieval tasks paradoxically degrades performance and exponentially inflates execution time (e.g., from 3.6s to 344.5s) and token costs. Code and data are available at: https://github.com/drbillwang/rlm-reproduction

Paperid: 275, https://arxiv.org/pdf/2603.02596.pdf GitHub

Abstract:
Tensegrity robots possess lightweight and resilient structures but present significant challenges for state estimation due to compliant and distributed ground contacts. This paper introduces a symmetry-aware heterogeneous graph neural network (Sym-HGNN) that infers contact states directly from proprioceptive measurements, including IMU and cable-length histories, without dedicated contact sensors. The network incorporates the robot's dihedral symmetry $D_3$ into the message-passing process to enhance sample efficiency and generalization. The predicted contacts are integrated into a state-of-the-art contact-aided invariant extended Kalman filter (InEKF) for improved pose estimation. Simulation results demonstrate that the proposed method achieves up to 15% higher accuracy and 5% higher F1-score using only 20% of the training data compared to the CNN and MI-HGNN baselines, while maintaining low-drift and physically consistent state estimation results comparable to ground truth contacts. This work highlights the potential of fully proprioceptive sensing for accurate and robust state estimation in tensegrity robots. Code available at: https://github.com/Jonathan-Twz/Tensegrity-Sym-HGNN

Paperid: 276, https://arxiv.org/pdf/2603.02573.pdf GitHub GitHub

Abstract:
Estimating the 3D trajectory of every pixel from a monocular video is crucial and promising for a comprehensive understanding of the 3D dynamics of videos. Recent monocular 3D tracking works demonstrate impressive performance, but are limited to either tracking sparse points on the first frame or a slow optimization-based framework for dense tracking. In this paper, we propose a feedforward model, called Track4World, enabling an efficient holistic 3D tracking of every pixel in the world-centric coordinate system. Built on the global 3D scene representation encoded by a VGGT-style ViT, Track4World applies a novel 3D correlation scheme to simultaneously estimate the pixel-wise 2D and 3D dense flow between arbitrary frame pairs. The estimated scene flow, along with the reconstructed 3D geometry, enables subsequent efficient 3D tracking of every pixel of this video. Extensive experiments on multiple benchmarks demonstrate that our approach consistently outperforms existing methods in 2D/3D flow estimation and 3D tracking, highlighting its robustness and scalability for real-world 4D reconstruction tasks.

Paperid: 277, https://arxiv.org/pdf/2603.02560.pdf GitHub

Abstract:
Multimodal Image Fusion (MMIF) integrates complementary information from various modalities to produce clearer and more informative fused images. MMIF under adverse weather is particularly crucial in autonomous driving and UAV monitoring applications. However, existing adverse weather fusion methods generally only tackle single types of degradation such as haze, rain, or snow, and fail when multiple degradations coexist (e.g., haze+rain, rain+snow). To address this challenge, we propose Compound Adverse Weather Mamba (CAWM-Mamba), the first end-to-end framework that jointly performs image fusion and compound weather restoration with unified shared weights. Our network contains three key components: (1) a Weather-Aware Preprocess Module (WAPM) to enhance degraded visible features and extracts global weather embeddings; (2) a Cross-modal Feature Interaction Module (CFIM) to facilitate the alignment of heterogeneous modalities and exchange of complementary features across modalities; and (3) a Wavelet Space State Block (WSSB) that leverages wavelet-domain decomposition to decouple multi-frequency degradations. WSSB includes Freq-SSM, a module that models anisotropic high-frequency degradation without redundancy, and a unified degradation representation mechanism to further improve generalization across complex compound weather conditions. Extensive experiments on the AWMM-100K benchmark and three standard fusion datasets demonstrate that CAWM-Mamba consistently outperforms state-of-the-art methods in both compound and single-weather scenarios. In addition, our fusion results excel in downstream tasks covering semantic segmentation and object detection, confirming the practical value in real-world adverse weather perception. The source code will be available at https://github.com/Feecuin/CAWM-Mamba.

Paperid: 278, https://arxiv.org/pdf/2603.02557.pdf GitHub

Abstract:
Vision-language models like CLIP have achieved remarkable progress in cross-modal representation learning, yet suffer from systematic misclassifications among visually and semantically similar categories. We observe that such confusion patterns are not random but persistently occur between specific category pairs, revealing the model's intrinsic bias and limited fine-grained discriminative ability. To address this, we propose CAPT, a Confusion-Aware Prompt Tuning framework that enables models to learn from their own misalignment. Specifically, we construct a Confusion Bank to explicitly model stable confusion relationships across categories and misclassified samples. On this basis, we introduce a Semantic Confusion Miner (SEM) to capture global inter-class confusion through semantic difference and commonality prompts, and a Sample Confusion Miner (SAM) to retrieve representative misclassified instances from the bank and capture sample-level cues through a Diff-Manner Adapter that integrates global and local contexts. To further unify confusion information across different granularities, a Multi-Granularity Difference Expert (MGDE) module is designed to jointly leverage semantic- and sample-level experts for more robust confusion-aware reasoning. Extensive experiments on 11 benchmark datasets demonstrate that our method significantly reduces confusion-induced errors while enhancing the discriminability and generalization of both base and novel classes, successfully resolving 50.72 percent of confusable sample pairs. Code will be released at https://github.com/greatest-gourmet/CAPT.

Paperid: 279, https://arxiv.org/pdf/2603.02556.pdf GitHub

Abstract:
Reasoning has emerged as a key capability of large language models. In linguistic tasks, this capability can be enhanced by self-improving techniques that refine reasoning paths for subsequent finetuning. However, extending these language-based self-improving approaches to vision language models (VLMs) presents a unique challenge:~visual hallucinations in reasoning paths cannot be effectively verified or rectified. Our solution starts with a key observation about visual contrast: when presented with a contrastive VQA pair, i.e., two visually similar images with synonymous questions, VLMs identify relevant visual cues more precisely. Motivated by this observation, we propose Visual Contrastive Self-Taught Reasoner (VC-STaR), a novel self-improving framework that leverages visual contrast to mitigate hallucinations in model-generated rationales. We collect a diverse suite of VQA datasets, curate contrastive pairs according to multi-modal similarity, and generate rationales using VC-STaR. Consequently, we obtain a new visual reasoning dataset, VisCoR-55K, which is then used to boost the reasoning capability of various VLMs through supervised finetuning. Extensive experiments show that VC-STaR not only outperforms existing self-improving approaches but also surpasses models finetuned on the SoTA visual reasoning datasets, demonstrating that the inherent contrastive ability of VLMs can bootstrap their own visual reasoning. Project at: https://github.com/zhiyupan42/VC-STaR.

Paperid: 280, https://arxiv.org/pdf/2603.02554.pdf GitHub

Abstract:
Knowledge distillation (KD) has been widely applied in semantic segmentation to compress large models, but conventional approaches primarily preserve in-domain accuracy while neglecting out-of-domain generalization, which is essential under distribution shifts. This limitation becomes more severe with the emergence of vision foundation models (VFMs): although VFMs exhibit strong robustness on unseen data, distilling them with conventional KD often compromises this ability. We propose Generalizable Knowledge Distillation (GKD), a multi-stage framework that explicitly enhances generalization. GKD decouples representation learning from task learning. In the first stage, the student acquires domain-agnostic representations through selective feature distillation, and in the second stage, these representations are frozen for task adaptation, thereby mitigating overfitting to visible domains. To further support transfer, we introduce a query-based soft distillation mechanism, where student features act as queries to teacher representations to selectively retrieve transferable spatial knowledge from VFMs. Extensive experiments on five domain generalization benchmarks demonstrate that GKD consistently outperforms existing KD methods, achieving average gains of +1.9% in foundation-to-foundation (F2F) and +10.6% in foundation-to-local (F2L) distillation. The code will be available at https://github.com/Younger-hua/GKD.

Paperid: 281, https://arxiv.org/pdf/2603.02553.pdf GitHub

Abstract:
During surgery, scrub nurses are required to frequently deliver surgical instruments to surgeons, which can lead to physical fatigue and decreased focus. Robotic scrub nurses provide a promising solution that can replace repetitive tasks and enhance efficiency. Existing research on robotic scrub nurses relies on predefined paths for instrument delivery, which limits their generalizability and poses safety risks in dynamic environments. To address these challenges, we present a collision-free dual-arm surgical assistive robot capable of performing instrument delivery. A vision-language model is utilized to automatically generate the robot's grasping and delivery trajectories in a zero-shot manner based on surgeons' instructions. A real-time obstacle minimum distance perception method is proposed and integrated into a unified quadratic programming framework. This framework ensures reactive obstacle avoidance and self-collision prevention during the dual-arm robot's autonomous movement in dynamic environments. Extensive experimental validations demonstrate that the proposed robotic system achieves an 83.33% success rate in surgical instrument delivery while maintaining smooth, collision-free movement throughout all trials. The project page and source code are available at https://give-me-scissors.github.io/.

Paperid: 282, https://arxiv.org/pdf/2603.02537.pdf GitHub

Abstract:
With the development of large language models (LLMs), numerous studies integrate LLMs through operator-like components to enhance relational data processing tasks, e.g., filters with semantic predicates, knowledge-augmented table imputation, reasoning-driven entity matching and more challenging semantic query processing. These components invoke LLMs while preserving a relational input/output interface, which we refer to as LLM-Enhanced Relational Operators (LROs). From an operator perspective, unfortunately, these existing LROs suffer from fragmented definition, various implementation strategies and inadequate evaluation benchmarks. To this end, in this paper, we first establish a unified LRO taxonomy to align existing LROs, and categorize them into: Select, Match, Impute, Cluster and Order, along with their operands and implementation variants. Second, we design LROBench, a comprehensive benchmark featuring 290 single-LRO queries and 60 multi-LRO queries, spanning 27 databases across more than 10 domains. LROBench covers all operating logics and operand granularities in its single-LRO workload, and provides challenging multi-LRO queries stratified by query complexity. Based on these, we evaluate individual LROs under various implementations, deriving practical insights into LRO design choices and summarizing our empirical best practices. We further compare the end-to-end performance of existing multi-LRO systems against an LRO suite instantiated with these best practices, in order to investigate how to design an effective LRO set for multi-LRO systems targeting complex semantic queries. Last, to facilitate future work, we outline promising future directions and open-source all benchmark data and evaluation code, available at https://github.com/LROBench/LROBench/.

Paperid: 283, https://arxiv.org/pdf/2603.02532.pdf GitHub

Abstract:
Multi-modal collaborative perception calls for great attention to enhancing the safety of autonomous driving. However, current multi-modal approaches remain a ``local fusion to communication'' sequence, which fuses multi-modal data locally and needs high bandwidth to transmit an individual's feature data before collaborative fusion. EIMC innovatively proposes an early collaborative paradigm. It injects lightweight collaborative voxels, transmitted by neighbor agents, into the ego's local modality-fusion step, yielding compact yet informative 3D collaborative priors that tighten cross-modal alignment. Next, a heatmap-driven consensus protocol identifies exactly where cooperation is needed by computing per-pixel confidence heatmaps. Only the Top-K instance vectors located in these low-confidence, high-discrepancy regions are queried from peers, then fused via cross-attention for completion. Afterwards, we apply a refinement fusion that involves collecting the top-K most confident instances from each agent and enhancing their features using self-attention. The above instance-centric messaging reduces redundancy while guaranteeing that critical occluded objects are recovered. Evaluated on OPV2V and DAIR-V2X, EIMC attains 73.01\% AP@0.5 while reducing byte bandwidth usage by 87.98\% compared with the best published multi-modal collaborative detector. Code publicly released at https://github.com/sidiangongyuan/EIMC.

Paperid: 284, https://arxiv.org/pdf/2603.02510.pdf GitHub

Abstract:
The transition from sequential to parallel computing is essential for modern high-performance applications but is hindered by the steep learning curve of concurrent programming. This challenge is magnified for irregular data structures (such as sparse graphs, unbalanced trees, and non-uniform meshes) where static scheduling fails and data dependencies are unpredictable. Current Large Language Models (LLMs) often fail catastrophically on these tasks, generating code plagued by subtle race conditions, deadlocks, and sub-optimal scaling. We bridge this gap with ParEVO, a framework designed to synthesize high-performance parallel algorithms for irregular data. Our contributions include: (1) The Parlay-Instruct Corpus, a curated dataset of 13,820 tasks synthesized via a "Critic-Refine" pipeline that explicitly filters for empirically performant algorithms that effectively utilize Work-Span parallel primitives; (2) specialized DeepSeek, Qwen, and Gemini models fine-tuned to align probabilistic generation with the rigorous semantics of the ParlayLib library; and (3) an Evolutionary Coding Agent (ECA) that improves the "last mile" of correctness by iteratively repairing code using feedback from compilers, dynamic race detectors, and performance profilers. On the ParEval benchmark, ParEVO achieves an average 106x speedup (with a maximum of 1103x) across the suite, and a robust 13.6x speedup specifically on complex irregular graph problems, outperforming state-of-the-art commercial models. Furthermore, our evolutionary approach matches state-of-the-art expert human baselines, achieving up to a 4.1x speedup on specific highly-irregular kernels. Source code and datasets are available at https://github.com/WildAlg/ParEVO.

Paperid: 285, https://arxiv.org/pdf/2603.02473.pdf GitHub

Abstract:
Memory-augmented LLM agents store and retrieve information from prior interactions, yet the relative importance of how memories are written versus how they are retrieved remains unclear. We introduce a diagnostic framework that analyzes how performance differences manifest across write strategies, retrieval methods, and memory utilization behavior, and apply it to a 3x3 study crossing three write strategies (raw chunks, Mem0-style fact extraction, MemGPT-style summarization) with three retrieval methods (cosine, BM25, hybrid reranking). On LoCoMo, retrieval method is the dominant factor: average accuracy spans 20 points across retrieval methods (57.1% to 77.2%) but only 3-8 points across write strategies. Raw chunked storage, which requires zero LLM calls, matches or outperforms expensive lossy alternatives, suggesting that current memory pipelines may discard useful context that downstream retrieval mechanisms fail to compensate for. Failure analysis shows that performance breakdowns most often manifest at the retrieval stage rather than at utilization. We argue that, under current retrieval practices, improving retrieval quality yields larger gains than increasing write-time sophistication. Code is publicly available at https://github.com/boqiny/memory-probe.

Paperid: 286, https://arxiv.org/pdf/2603.02462.pdf GitHub

Abstract:
A key challenge in deriving unified neural solvers for combinatorial optimization (CO) is efficient generalization of models between a given set of tasks to new tasks not used during the initial training process. To address it, we first establish a new model, which uses a GCON module as a form of expressive message passing together with energy-based unsupervised loss functions. This model achieves high performance (often comparable with state-of-the-art results) across multiple CO tasks when trained individually on each task. We then leverage knowledge from the computational reducibility literature to propose pretraining and fine-tuning strategies that transfer effectively (a) between MVC, MIS and MaxClique, and (b) in a multi-task learning setting that additionally incorporates MaxCut, MDS and graph coloring. Additionally, in a leave-one-out, multi-task learning setting, we observe that pretraining on all but one task almost always leads to faster convergence on the remaining task when fine-tuning while avoiding negative transfer. Our findings indicate that learning common representations across multiple graph CO problems is viable through the use of expressive message passing coupled with pretraining strategies that are informed by the polynomial reduction literature, thereby taking an important step towards enabling the development of foundational models for neural CO. We provide an open-source implementation of our work at https://github.com/semihcanturk/COPT-MT .

Paperid: 287, https://arxiv.org/pdf/2603.02406.pdf GitHub

Abstract:
Generative models have recently advanced $\textit{de novo}$ protein design by learning the statistical regularities of natural structures. However, current approaches face three key limitations: (1) Existing methods cannot jointly learn protein geometry and design tasks, where pretraining can be a solution; (2) Current pretraining methods mostly rely on local, non-rigid atomic representations for property prediction downstream tasks, limiting global geometric understanding for protein generation tasks; and (3) Existing approaches have yet to effectively model the rich dynamic and conformational information of protein structures. To overcome these issues, we introduce $\textbf{RigidSSL}$ ($\textit{Rigidity-Aware Self-Supervised Learning}$), a geometric pretraining framework that front-loads geometry learning prior to generative finetuning. Phase I (RigidSSL-Perturb) learns geometric priors from 432K structures from the AlphaFold Protein Structure Database with simulated perturbations. Phase II (RigidSSL-MD) refines these representations on 1.3K molecular dynamics trajectories to capture physically realistic transitions. Underpinning both phases is a bi-directional, rigidity-aware flow matching objective that jointly optimizes translational and rotational dynamics to maximize mutual information between conformations. Empirically, RigidSSL variants improve designability by up to 43% while enhancing novelty and diversity in unconditional generation. Furthermore, RigidSSL-Perturb improves the success rate by 5.8% in zero-shot motif scaffolding and RigidSSL-MD captures more biophysically realistic conformational ensembles in G protein-coupled receptor modeling.

Paperid: 288, https://arxiv.org/pdf/2603.02364.pdf GitHub GitHub

Abstract:
We introduce LRLspoof, a large-scale multilingual synthetic-speech corpus for cross-lingual spoof detection, comprising 2,732 hours of audio generated with 24 open-source TTS systems across 66 languages, including 45 low-resource languages under our operational definition. To evaluate robustness without requiring target-domain bonafide speech, we benchmark 11 publicly available countermeasures using threshold transfer: for each model we calibrate an EER operating point on pooled external benchmarks and apply the resulting threshold, reporting spoof rejection rate (SRR). Results show model-dependent cross-lingual disparity, with spoof rejection varying markedly across languages even under controlled conditions, highlighting language as an independent source of domain shift in spoof detection. The dataset is publicly available at \href{https://huggingface.co/datasets/MTUCI/LRLspoof}{\textbf{\underline{\textit{HuggingFace}}}} and \href{https://modelscope.cn/datasets/lab260/LRLspoof}{\textbf{\underline{\textit{ModelScope}}}}

Paperid: 289, https://arxiv.org/pdf/2603.02363.pdf GitHub

Abstract:
In this work, we investigate the degradation of existing VMR methods, particularly of DETR architectures, when trained on caption-based queries but evaluated on search queries. For this, we introduce three benchmarks by modifying the textual queries in three public VMR datasets -- i.e., HD-EPIC, YouCook2 and ActivityNet-Captions. Our analysis reveals two key generalization challenges: (i) A language gap, arising from the linguistic under-specification of search queries, and (ii) a multi-moment gap, caused by the shift from single-moment to multi-moment queries. We also identify a critical issue in these architectures -- an active decoder-query collapse -- as a primary cause of the poor generalization to multi-moment instances. We mitigate this issue with architectural modifications that effectively increase the number of active decoder queries. Extensive experiments demonstrate that our approach improves performance on search queries by up to 14.82% mAP_m, and up to 21.83% mAP_m on multi-moment search queries. The code, models and data are available in the project webpage: https://davidpujol.github.io/beyond-vmr/

Paperid: 290, https://arxiv.org/pdf/2603.02351.pdf GitHub

Abstract:
Recent advancements in neural visual geometry, including transformer-based models such as VGGT and Pi3, have achieved impressive accuracy on 3D reconstruction tasks. However, their reliance on full attention makes them fundamentally limited by GPU memory capacity, preventing them from scaling to large, unordered image collections. We introduce MERG3R, a training-free divide-and-conquer framework that enables geometric foundation models to operate far beyond their native memory limits. MERG3R first reorders and partitions unordered images into overlapping, geometrically diverse subsets that can be reconstructed independently. It then merges the resulting local reconstructions through an efficient global alignment and confidence-weighted bundle adjustment procedure, producing a globally consistent 3D model. Our framework is model-agnostic and can be paired with existing neural geometry models. Across large-scale datasets, including 7-Scenes, NRGBD, Tanks & Temples, and Cambridge Landmarks, MERG3R consistently improves reconstruction accuracy, memory efficiency, and scalability, enabling high-quality reconstruction when the dataset exceeds memory capacity limits.

Paperid: 291, https://arxiv.org/pdf/2603.02331.pdf GitHub

Abstract:
We develop a flexible neural demand system for continuous budget allocation that estimates budget shares on the simplex by minimizing KL divergence. Shares are produced via a softmax of a state-dependent preference scorer and disciplined with regularity penalties (monotonicity, Slutsky symmetry) to support coherent comparative statics and welfare without imposing a parametric utility form. State dependence enters through a habit stock defined as an exponentially weighted moving average of past consumption. Simulations recover elasticities and welfare accurately and show sizable gains when habit formation is present. In our empirical application using Dominick's analgesics data, adding habit reduces out-of-sample error by c.33%, reshapes substitution patterns, and increases CV losses from a 10% ibuprofen price rise by about 15-16% relative to a static model. The code is available at https://github.com/martagrz/neural_demand_habit .

Paperid: 292, https://arxiv.org/pdf/2603.02329.pdf GitHub

Abstract:
Humans commonly identify 3D object affordance through observed interactions in images or videos, and once formed, such knowledge can be generically generalized to novel objects. Inspired by this principle, we advocate for a novel framework that leverages emerging multimodal large language models (MLLMs) for interaction intention-driven 3D affordance grounding, namely HAMMER. Instead of generating explicit object attribute descriptions or relying on off-the-shelf 2D segmenters, we alternatively aggregate the interaction intention depicted in the image into a contact-aware embedding and guide the model to infer textual affordance labels, ensuring it thoroughly excavates object semantics and contextual cues. We further devise a hierarchical cross-modal integration mechanism to fully exploit the complementary information from the MLLM for 3D representation refinement and introduce a multi-granular geometry lifting module that infuses spatial characteristics into the extracted intention embedding, thus facilitating accurate 3D affordance localization. Extensive experiments on public datasets and our newly constructed corrupted benchmark demonstrate the superiority and robustness of HAMMER compared to existing approaches. All code and weights are publicly available.

Paperid: 293, https://arxiv.org/pdf/2603.02294.pdf GitHub

Abstract:
Long-tailed class distributions pose a significant challenge for multi-label chest X-ray (CXR) classification, where rare but clinically important findings are severely underrepresented. In this work, we present a systematic empirical evaluation of loss functions, CNN backbone architectures and post-training strategies on the CXR-LT 2026 benchmark, comprising approximately 143K images with 30 disease labels from PadChest. Our experiments demonstrate that LDAM with deferred re-weighting (LDAM-DRW) consistently outperforms standard BCE and asymmetric losses for rare class recognition. Amongst the architectures evaluated, ConvNeXt-Large achieves the best single-model performance with 0.5220 mAP and 0.3765 F1 on our development set, whilst classifier re-training and test-time augmentation further improve ranking metrics. On the official test leaderboard, our submission achieved 0.3950 mAP, ranking 5th amongst all 68 participating teams with total of 1528 submissions. We provide a candid analysis of the development-to-test performance gap and discuss practical insights for handling class imbalance in clinical imaging settings. Code is available at https://github.com/Nikhil-Rao20/Long_Tail.

Paperid: 294, https://arxiv.org/pdf/2603.02288.pdf GitHub

Abstract:
Facial feminization surgery (FFS) is a key component of gender affirmation for transgender and gender diverse patients, aiming to reshape craniofacial structures toward a female morphology. Current surgical planning procedures largely rely on subjective clinical assessment, lacking quantitative and reproducible anatomical guidance. We therefore propose AutoFFS, a novel data-driven framework that generates counterfactual skull morphologies through adversarial free-form deformations. Our method performs a deformation-based targeted adversarial attack on an ensemble of pre-trained binary sex classifiers that learned sexual dimorphism, effectively transforming individual skull shapes toward the target sex. The generated counterfactual skull morphologies provide a quantitative foundation for preoperative planning in FFS, driving advances in this largely overlooked patient group. We validate our approach through classifier-based evaluation and a human perceptual study, confirming that the generated morphologies exhibit target sex characteristics.

Paperid: 295, https://arxiv.org/pdf/2603.02286.pdf GitHub

Abstract:
Incremental Object Detection (IOD) aims to continuously learn new object categories without forgetting previously learned ones. Recently, prompt-based methods have gained popularity for their replay-free design and parameter efficiency. However, due to prompt coupling and prompt drift, these methods often suffer from prompt degradation during continual adaptation. To address these issues, we propose a novel prompt-decoupled framework called PDP. PDP innovatively designs a dual-pool prompt decoupling paradigm, which consists of a shared pool used to capture task-general knowledge for forward transfer, and a private pool used to learn task-specific discriminative features. This paradigm explicitly separates task-general and task-specific prompts, preventing interference between prompts and mitigating prompt coupling. In addition, to counteract prompt drift resulting from inconsistent supervision where old foreground objects are treated as background in subsequent tasks, PDP introduces a Prototypical Pseudo-Label Generation (PPG) module. PPG can dynamically update the class prototype space during training and use the class prototypes to further filter valuable pseudo-labels, maintaining supervisory signal consistency throughout the incremental process. PDP achieves state-of-the-art performance on MS-COCO (with a 9.2\% AP improvement) and PASCAL VOC (with a 3.3\% AP improvement) benchmarks, highlighting its potential in balancing stability and plasticity. The code and dataset are released at: https://github.com/zyt95579/PDP\_IOD/tree/main

Paperid: 296, https://arxiv.org/pdf/2603.02258.pdf GitHub

Abstract:
Do neural machine translation models learn language-universal conceptual representations, or do they merely cluster languages by surface similarity? We investigate this question by probing the representation geometry of Meta's NLLB-200, a 200-language encoder-decoder Transformer, through six experiments that bridge NLP interpretability with cognitive science theories of multilingual lexical organization. Using the Swadesh core vocabulary list embedded across 135 languages, we find that the model's embedding distances significantly correlate with phylogenetic distances from the Automated Similarity Judgment Program ($ρ= 0.13$, $p = 0.020$), demonstrating that NLLB-200 has implicitly learned the genealogical structure of human languages. We show that frequently colexified concept pairs from the CLICS database exhibit significantly higher embedding similarity than non-colexified pairs ($U = 42656$, $p = 1.33 \times 10^{-11}$, $d = 0.96$), indicating that the model has internalized universal conceptual associations. Per-language mean-centering of embeddings improves the between-concept to within-concept distance ratio by a factor of 1.19, providing geometric evidence for a language-neutral conceptual store analogous to the anterior temporal lobe hub identified in bilingual neuroimaging. Semantic offset vectors between fundamental concept pairs (e.g., man to woman, big to small) show high cross-lingual consistency (mean cosine = 0.84), suggesting that second-order relational structure is preserved across typologically diverse languages. We release InterpretCognates, an open-source interactive toolkit for exploring these phenomena, alongside a fully reproducible analysis pipeline.

Paperid: 297, https://arxiv.org/pdf/2603.02241.pdf GitHub

Abstract:
This study establishes a benchmark for Caenorhabditis elegans neuron classification, comparing four graph methods (GCN, GraphSAGE, GAT, GraphTransformer) against four non-graph methods (Logistic Regression, MLP, LOLCAT, NeuPRINT). Using the functional connectome, we classified Sensory, Interneuron, and Motor neurons based on Spatial, Connection, and Neuronal Activity features. Results show that attention-based GNNs significantly outperform baselines on the Spatial and Connection features. The Neuronal Activity features yielded poor performance, likely due to the low temporal resolution of the underlying neuronal activity data. Our benchmark validates the use of GNNs and highlights that Spatial and Connection features are key predictors for Caenorhabditis elegans neuron classes. Code is available at: https://github.com/JingqiLuu/neuronclf-gnn-benchmark.

Paperid: 298, https://arxiv.org/pdf/2603.02240.pdf GitHub

Abstract:
We present SuperLocalMemory, a local-first memory system for multi-agent AI that defends against OWASP ASI06 memory poisoning through architectural isolation and Bayesian trust scoring, while personalizing retrieval through adaptive learning-to-rank -- all without cloud dependencies or LLM inference calls. As AI agents increasingly rely on persistent memory, cloud-based memory systems create centralized attack surfaces where poisoned memories propagate across sessions and users -- a threat demonstrated in documented attacks against production systems. Our architecture combines SQLite-backed storage with FTS5 full-text search, Leiden-based knowledge graph clustering, an event-driven coordination layer with per-agent provenance, and an adaptive re-ranking framework that learns user preferences through three-layer behavioral analysis (cross-project technology preferences, project context detection, and workflow pattern mining). Evaluation across seven benchmark dimensions demonstrates 10.6ms median search latency, zero concurrency errors under 10 simultaneous agents, trust separation (gap =0.90) with 72% trust degradation for sleeper attacks, and 104% improvement in NDCG@5 when adaptive re-ranking is enabled. Behavioral data is isolated in a separate database with GDPR Article 17 erasure support. SuperLocalMemory is open-source (MIT) and integrates with 17+ development tools via Model Context Protocol.

Paperid: 299, https://arxiv.org/pdf/2603.02236.pdf GitHub

Abstract:
Recent studies have demonstrated the potential of Large Language Models (LLMs) in generating GPU Kernels. Current benchmarks focus on the translation of high-level languages into CUDA, overlooking the more general and challenging task of text-to-CUDA generation. Furthermore, given the hardware-specific and performance-critical features of GPU programming, accurately assessing the performance of LLM-generated GPU programs is nontrivial. In this work, we introduce CUDABench, a comprehensive benchmark designed to evaluate the text-to-CUDA capabilities of LLMs. First, we construct CUDABench-Set, which covers Breadth-Depth-Difficulty evaluation space in diverse application domains, including artificial intelligence, scientific computing, and data analytics, etc. Furthermore, we propose CUDABench-Score and Generative Verification Pipeline that assess (1) compilation correctness, (2) functional consistency through execution-based verification, and (3) a novel roofline-based metric, Performance-Score. Benchmarking state-of-the-art LLMs reveals insightful findings and challenges of text-to-CUDA, such as a notable mismatch between high compilation success rates and low functional correctness, a lack of domain-specific algorithmic knowledge, and suboptimal utilization of GPU hardware resources. Our benchmark is available at https://github.com/CUDA-Bench/CUDABench.

Paperid: 300, https://arxiv.org/pdf/2603.02215.pdf GitHub

Abstract:
Chemical reaction prediction is pivotal for accelerating drug discovery and synthesis planning. Despite advances in data-driven models, current approaches are hindered by an overemphasis on parameter and dataset scaling. Some methods coupled with evaluation techniques that bypass fundamental challenges in reaction representation and fail to capture deep chemical intuition like reaction common sense and {topological atom mapping logic}. We argue that the core challenge lies in instilling these knowledge into the models. To this end, we propose a unified framework that prioritizes chemical understanding over scale through three key innovations: (1) a {Latent Chemical Consistency} objective that models reactions as movements on a continuous chemical manifold, ensuring reversible and physically plausible transformations; (2) a {Hierarchical Cognitive Curriculum} that trains the model through progressive stages, from syntax mastery to semantic reasoning, building robust chemical intuition; (3) {Atom-Map Permutation Invariance (AMPI)}, which force the model to learn invariant relational topology and balance multi-task learning. (4)and structured plan-based reasoning to improve the performance of the LLMs. Our compact {0.5B-parameter model}, \textbf{RxnNano} significantly outperforms fine-tuned LLMs ten times larger (>7B) and all the domain baselines, achieving a 23.5\% Top-1 accuracy improvement on rigorous benchmarks without test-time augmentation. https://github.com/rlisml/RxnNano.

Paperid: 301, https://arxiv.org/pdf/2603.02210.pdf GitHub

Abstract:
Human-product images, which showcase the integration of humans and products, play a vital role in advertising, e-commerce, and digital marketing. The essential challenge of generating such images lies in ensuring the high-fidelity preservation of product details. Among existing paradigms, reference-based inpainting offers a targeted solution by leveraging product reference images to guide the inpainting process. However, limitations remain in three key aspects: the lack of diverse large-scale training data, the struggle of current models to focus on product detail preservation, and the inability of coarse supervision for achieving precise guidance. To address these issues, we propose HiFi-Inpaint, a novel high-fidelity reference-based inpainting framework tailored for generating human-product images. HiFi-Inpaint introduces Shared Enhancement Attention (SEA) to refine fine-grained product features and Detail-Aware Loss (DAL) to enforce precise pixel-level supervision using high-frequency maps. Additionally, we construct a new dataset, HP-Image-40K, with samples curated from self-synthesis data and processed with automatic filtering. Experimental results show that HiFi-Inpaint achieves state-of-the-art performance, delivering detail-preserving human-product images.

Paperid: 302, https://arxiv.org/pdf/2603.02175.pdf GitHub GitHub

Abstract:
Instruction-based video editing has witnessed rapid progress, yet current methods often struggle with precise visual control, as natural language is inherently limited in describing complex visual nuances. Although reference-guided editing offers a robust solution, its potential is currently bottlenecked by the scarcity of high-quality paired training data. To bridge this gap, we introduce a scalable data generation pipeline that transforms existing video editing pairs into high-fidelity training quadruplets, leveraging image generative models to create synthesized reference scaffolds. Using this pipeline, we construct RefVIE, a large-scale dataset tailored for instruction-reference-following tasks, and establish RefVIE-Bench for comprehensive evaluation. Furthermore, we propose a unified editing architecture, Kiwi-Edit, that synergizes learnable queries and latent visual features for reference semantic guidance. Our model achieves significant gains in instruction following and reference fidelity via a progressive multi-stage training curriculum. Extensive experiments demonstrate that our data and architecture establish a new state-of-the-art in controllable video editing. All datasets, models, and code is released at https://github.com/showlab/Kiwi-Edit.

Paperid: 303, https://arxiv.org/pdf/2603.02134.pdf GitHub

Abstract:
Recent advances in generalizable 3D Gaussian Splatting (3DGS) have enabled rapid 3D scene reconstruction within seconds, eliminating the need for per-scene optimization. However, existing methods primarily follow an offline reconstruction paradigm, lacking the capacity for continuous reconstruction, which limits their applicability to online scenarios such as robotics and VR/AR. In this paper, we introduce OnlineX, a feed-forward framework that reconstructs both 3D visual appearance and language fields in an online manner using only streaming images. A key challenge in online formulation is the cumulative drift issue, which is rooted in the fundamental conflict between two opposing roles of the memory state: an active role that constantly refreshes to capture high-frequency local geometry, and a stable role that conservatively accumulates and preserves the long-term global structure. To address this, we introduce a decoupled active-to-stable state evolution paradigm. Our framework decouples the memory state into a dedicated active state and a persistent stable state, and then cohesively fuses the information from the former into the latter to achieve both fidelity and stability. Moreover, we jointly model visual appearance and language fields and incorporate an implicit Gaussian fusion module to enhance reconstruction quality. Experiments on mainstream datasets demonstrate that our method consistently outperforms prior work in novel view synthesis and semantic understanding, showcasing robust performance across input sequences of varying lengths with real-time inference speed.

Paperid: 304, https://arxiv.org/pdf/2603.02133.pdf GitHub

Abstract:
Compositional scene reconstruction seeks to create object-centric representations rather than holistic scenes from real-world videos, which is natively applicable for simulation and interaction. Conventional compositional reconstruction approaches primarily emphasize on visual appearance and show limited generalization ability to real-world scenarios. In this paper, we propose SimRecon, a framework that realizes a "Perception-Generation-Simulation" pipeline towards cluttered scene reconstruction, which first conducts scene-level semantic reconstruction from video input, then performs single-object generation, and finally assembles these assets in the simulator. However, naively combining these three stages leads to visual infidelity of generated assets and physical implausibility of the final scene, a problem particularly severe for complex scenes. Thus, we further propose two bridging modules between the three stages to address this problem. To be specific, for the transition from Perception to Generation, critical for visual fidelity, we introduce Active Viewpoint Optimization, which actively searches in 3D space to acquire optimal projected images as conditions for single-object completion. Moreover, for the transition from Generation to Simulation, essential for physical plausibility, we propose a Scene Graph Synthesizer, which guides the construction from scratch in 3D simulators, mirroring the native, constructive principle of the real world. Extensive experiments on the ScanNet dataset validate our method's superior performance over previous state-of-the-art approaches.

Paperid: 305, https://arxiv.org/pdf/2603.02099.pdf GitHub

Abstract:
Think-Answer reasoners such as DeepSeek-R1 have made notable progress by leveraging interpretable internal reasoning. However, despite the frequent presence of self-reflective cues like "Oops!", they remain vulnerable to output errors during single-pass inference. To address this limitation, we propose an efficient Recursive Think-Answer Process (R-TAP) that enables models to engage in iterative reasoning cycles and generate more accurate answers, going beyond conventional single-pass approaches. Central to this approach is a confidence generator that evaluates the certainty of model responses and guides subsequent improvements. By incorporating two complementary rewards-Recursively Confidence Increase Reward and Final Answer Confidence Reward-we show that R-TAP-enhanced models consistently outperform conventional single-pass methods for both large language models (LLMs) and vision-language models (VLMs). Moreover, by analyzing the frequency of "Oops"-like expressions in model responses, we find that R-TAP-applied models exhibit significantly fewer self-reflective patterns, resulting in more stable and faster inference-time reasoning. We hope R-TAP pave the way evolving into efficient and elaborated methods to refine the reasoning processes of future AI.

Paperid: 306, https://arxiv.org/pdf/2603.01741.pdf GitHub

Abstract:
Scaling reinforcement learning to tens of thousands of parallel environments requires overcoming the limited exploration capacity of a single policy. Ensemble-based policy gradient methods, which employ multiple policies to collect diverse samples, have recently been proposed to promote exploration. However, merely broadening the exploration space does not always enhance learning capability, since excessive exploration can reduce exploration quality or compromise training stability. In this work, we theoretically analyze the impact of inter-policy diversity on learning efficiency in policy ensembles, and propose Coupled Policy Optimization which regulates diversity through KL constraints between policies. The proposed method enables effective exploration and outperforms strong baselines such as SAPG, PBT, and PPO across multiple tasks, including challenging dexterous manipulation, in terms of both sample efficiency and final performance. Furthermore, analysis of policy diversity and effective sample size during training reveals that follower policies naturally distribute around the leader, demonstrating the emergence of structured and efficient exploratory behavior. Our results indicate that diverse exploration under appropriate regulation is key to achieving stable and sample-efficient learning in ensemble policy gradient methods. Project page at https://naoki04.github.io/paper-cpo/ .

Paperid: 307, https://arxiv.org/pdf/2603.01412.pdf GitHub

Abstract:
With growing real-world demands, efficient tracking has received increasing attention. However, most existing methods are limited to RGB inputs and struggle in multi-modal scenarios. Moreover, current multi-modal tracking approaches typically use complex designs, making them too heavy and slow for resource-constrained deployment. To tackle these limitations, we propose UETrack, an efficient framework for single object tracking. UETrack demonstrates high practicality and versatility, efficiently handling multiple modalities including RGB, Depth, Thermal, Event, and Language, and addresses the gap in efficient multi-modal tracking. It introduces two key components: a Token-Pooling-based Mixture-of-Experts mechanism that enhances modeling capacity through feature aggregation and expert specialization, and a Target-aware Adaptive Distillation strategy that selectively performs distillation based on sample characteristics, reducing redundant supervision and improving performance. Extensive experiments on 12 benchmarks across 3 hardware platforms show that UETrack achieves a superior speed-accuracy trade-off compared to previous methods. For instance, UETrack-B achieves 69.2% AUC on LaSOT and runs at 163/56/60 FPS on GPU/CPU/AGX, demonstrating strong practicality and versatility. Code is available at https://github.com/kangben258/UETrack.

Paperid: 308, https://arxiv.org/pdf/2603.01410.pdf GitHub

Abstract:
Knowledge graphs provide structured and reliable information for many real-world applications, motivating increasing interest in combining large language models (LLMs) with graph-based retrieval to improve factual grounding. Recent Graph-based Retrieval-Augmented Generation (GraphRAG) methods therefore introduce iterative interaction between LLMs and knowledge graphs to enhance reasoning capability. However, existing approaches typically depend on manually designed guidance and interact with knowledge graphs through a limited set of predefined tools, which substantially constrains graph exploration. To address these limitations, we propose GraphScout, a training-centric agentic graph reasoning framework equipped with more flexible graph exploration tools. GraphScout enables models to autonomously interact with knowledge graphs to synthesize structured training data which are then used to post-train LLMs, thereby internalizing agentic graph reasoning ability without laborious manual annotation or task curation. Extensive experiments across five knowledge-graph domains show that a small model (e.g., Qwen3-4B) augmented with GraphScout outperforms baseline methods built on leading LLMs (e.g., Qwen-Max) by an average of 16.7\% while requiring significantly fewer inference tokens. Moreover, GraphScout exhibits robust cross-domain transfer performance. Our code will be made publicly available~\footnote{https://github.com/Ying-Yuchen/_GraphScout_}.

Paperid: 309, https://arxiv.org/pdf/2603.01400.pdf GitHub GitHub

Abstract:
Video Large Language Models (VLLMs) demonstrate strong video understanding but suffer from inefficiency due to redundant visual tokens. Existing pruning primary targets intra-frame spatial redundancy or prunes inside the LLM with shallow-layer overhead, yielding suboptimal spatiotemporal reduction and underutilizing long-context compressibility. All of them often discard subtle yet informative context from merged or pruned tokens. In this paper, we propose a new perspective that elaborates token \textbf{A}nchors within intra-frame and inter-frame to comprehensively aggregate the informative contexts via local-global \textbf{O}ptimal \textbf{T}ransport (\textbf{AOT}). Specifically, we first establish local- and global-aware token anchors within each frame under the attention guidance, which then optimal transport aggregates the informative contexts from pruned tokens, constructing intra-frame token anchors. Then, building on the temporal frame clips, the first frame within each clip will be considered as the keyframe anchors to ensemble similar information from consecutive frames through optimal transport, while keeping distinct tokens to represent temporal dynamics, leading to efficient token reduction in a training-free manner. Extensive evaluations show that our proposed AOT obtains competitive performances across various short- and long-video benchmarks on leading video LLMs, obtaining substantial computational efficiency while preserving temporal and visual fidelity. Project webpage: \href{https://tyroneli.github.io/AOT}{AOT}.

Paperid: 310, https://arxiv.org/pdf/2603.01399.pdf GitHub

Abstract:
Speculative Decoding (SD) has emerged as a premier technique for accelerating Large Language Model (LLM) inference by decoupling token generation into rapid drafting and parallel verification. While recent advancements in self-speculation and lookahead decoding have successfully minimized drafting overhead, they have shifted the primary performance bottleneck to the verification phase. Since verification requires a full forward pass of the target model, it remains strictly memory-bandwidth bound, fundamentally limiting the maximum achievable speedup.In this paper, we introduce \textbf{Quasar} (\textbf{Qua}ntized \textbf{S}elf-speculative \textbf{A}cceleration for \textbf{R}apid Inference), a novel, training-free framework designed to overcome this "memory wall" by employing low-bit quantization specifically for the verification stage. Our empirical analysis reveals that while aggressive structural pruning significantly degrades verification accuracy, quantization-based verification preserves the logit distribution with high fidelity while effectively halving memory traffic. Extensive experiments on state-of-the-art models (e.g., OpenPangu and Qwen3) demonstrate that Quasar maintains a speculative acceptance length comparable to full-precision methods while achieving a $1.28\times$ improvement in end-to-end throughput. Being orthogonal to existing drafting strategies, Quasar offers a generic and efficient pathway to accelerate the verification leg of speculative execution. Code is available at https://github.com/Tom-HG/Quasar.

Paperid: 311, https://arxiv.org/pdf/2603.01376.pdf GitHub

Abstract:
Sparse plus Low-Rank $(\mathbf{S} + \mathbf{LR})$ decomposition of Large Language Models (LLMs) has emerged as a promising direction in model compression, aiming to decompose pre-trained model weights into a sum of sparse and low-rank matrices $(\mathbf{W} \approx \mathbf{S} + \mathbf{LR})$. Despite recent progress, existing methods often suffer from substantial performance degradation compared to dense models. In this work, we introduce 3BASiL-TM, an efficient one-shot post-training method for $(\mathbf{S} + \mathbf{LR})$ decomposition of LLMs that addresses this gap. Our approach first introduces a novel 3-Block Alternating Direction Method of Multipliers (ADMM) method, termed 3BASiL, to minimize the layer-wise reconstruction error with convergence guarantees. We then design an efficient transformer-matching (TM) refinement step that jointly optimizes the sparse and low-rank components across transformer layers. This step minimizes a novel memory-efficient loss that aligns outputs at the transformer level. Notably, the TM procedure is universal as it can enhance any $(\mathbf{S} + \mathbf{LR})$ decomposition, including pure sparsity. Our numerical experiments show that 3BASiL-TM reduces the WikiText2 perplexity gap relative to dense LLaMA-8B model by over 30% under a (2:4 Sparse + 64 LR) configuration, compared to prior methods. Moreover, our method achieves over 2.5x faster compression runtime on an A100 GPU compared to SOTA $(\mathbf{S} + \mathbf{LR})$ method. Our code is available at https://github.com/mazumder-lab/3BASiL.

Paperid: 312, https://arxiv.org/pdf/2603.01361.pdf GitHub

Abstract:
Feature encoders play a key role in pixel-level crack segmentation by shaping the representation of fine textures and thin structures. Existing CNN-, Transformer-, and Mamba-based models each capture only part of the required spatial or structural information, leaving clear gaps in modeling complex crack patterns. To address this, we present MixerCSeg, a mixer architecture designed like a coordinated team of specialists, where CNN-like pathways focus on local textures, Transformer-style paths capture global dependencies, and Mamba-inspired flows model sequential context within a single encoder. At the core of MixerCSeg is the TransMixer, which explores Mamba's latent attention behavior while establishing dedicated pathways that naturally express both locality and global awareness. To further enhance structural fidelity, we introduce a spatial block processing strategy and a Direction-guided Edge Gated Convolution (DEGConv) that strengthens edge sensitivity under irregular crack geometries with minimal computational overhead. A Spatial Refinement Multi-Level Fusion (SRF) module is then employed to refine multi-scale details without increasing complexity. Extensive experiments on multiple crack segmentation benchmarks show that MixerCSeg achieves state-of-the-art performance with only 2.05 GFLOPs and 2.54 M parameters, demonstrating both efficiency and strong representational capability. The code is available at https://github.com/spiderforest/MixerCSeg.

Paperid: 313, https://arxiv.org/pdf/2603.01345.pdf GitHub

Abstract:
Multi-objective optimization is now a core paradigm in engineering design and scientific discovery. Yet mainstream evolutionary frameworks, including \textit{pymoo}, still depend on imperative coding for problem definition, algorithm configuration, and post-hoc analysis. That requirement creates a non-trivial barrier for practitioners without strong software-engineering training and often complicates reproducible experimentation. We address this gap through PymooLab, an open-source visual analytics environment built on top of \textit{pymoo}. The platform unifies configuration, execution monitoring, and formal decision support in a single reproducible workflow that automatically records hyperparameters, evaluation budgets, and random seeds. Its decoupled object-oriented architecture preserves compatibility with the base ecosystem while enabling LLM-assisted code generation for rapid model formulation. The interface also embeds interactive Multi-Criteria Decision Making (MCDM) tools, which reduces the cognitive burden of Pareto-front inspection. For computationally intensive studies, PymooLab relies on the native \textit{pymoo} acceleration pathway through JAX, improving scalability in high-dimensional evaluations. Overall, the framework combines visual experimentation, LLM-based modeling, and deterministic orchestration to narrow the gap between rigorous operations research and practical accessibility for domain experts. Source code is publicly available at https://github.com/METISBR/pymoolab.

Paperid: 314, https://arxiv.org/pdf/2603.01295.pdf GitHub GitHub

Abstract:
Breast ultrasound interpretation requires simultaneous lesion segmentation and tissue classification. However, conventional multi-task learning approaches suffer from task interference and rigid coordination strategies that fail to adapt to instance-specific prediction difficulty. We propose a multi-task framework addressing these limitations through multi-level decoder interaction and uncertainty-aware adaptive coordination. Task Interaction Modules operate at all decoder levels, establishing bidirectional segmentation-classification communication during spatial reconstruction through attention weighted pooling and multiplicative modulation. Unlike prior single-level or encoder-only approaches, this multi-level design captures scale specific task synergies across semantic-to-spatial scales, producing complementary task interaction streams. Uncertainty-Proxy Attention adaptively weights base versus enhanced features at each level using feature activation variance, enabling per-level and per-sample task balancing without heuristic tuning. To support instance-adaptive prediction, multi-scale context fusion captures morphological cues across varying lesion sizes. Evaluation on multiple publicly available breast ultrasound datasets demonstrates competitive performance, including 74.5% lesion IoU and 90.6% classification accuracy on BUSI dataset. Ablation studies confirm that multi-level task interaction provides significant performance gains, validating that decoder-level bidirectional communication is more effective than conventional encoder-only parameter sharing. The code is available at: https://github.com/C-loud-Nine/Uncertainty-Aware-Multi-Level-Decoder-Interaction.

Paperid: 315, https://arxiv.org/pdf/2603.01292.pdf GitHub

Abstract:
This paper proposes Proximal Policy Optimization with Linear Temporal Logic Constraints (PPO-LTL), a framework that integrates safety constraints written in LTL into PPO for safe reinforcement learning. LTL constraints offer rigorous representations of complex safety requirements, such as regulations that broadly exist in robotics, enabling systematic monitoring of safety requirements. Violations against LTL constraints are monitored by limit-deterministic Büchi automata, and then translated by a logic-to-cost mechanism into penalty signals. The signals are further employed for guiding the policy optimization via the Lagrangian scheme. Extensive experiments on the Zones and CARLA environments show that our PPO-LTL can consistently reduce safety violations, while maintaining competitive performance, against the state-of-the-art methods. The code is at https://github.com/EVIEHub/PPO-LTL.

Paperid: 316, https://arxiv.org/pdf/2603.01291.pdf GitHub

Abstract:
Fake news undermines societal trust and decision-making across politics, economics, health, and international relations, and in extreme cases threatens human lives and societal safety. Because fake news reflects region-specific political, social, and cultural contexts and is expressed in language, evaluating the risks of large language models (LLMs) requires a multi-lingual and regional perspective. Malicious users can bypass safeguards through jailbreak attacks, inducing LLMs to generate fake news. However, no benchmark currently exists to systematically assess attack resilience across languages and regions. Here, we propose JailNewsBench, the first benchmark for evaluating LLM robustness against jailbreak-induced fake news generation. JailNewsBench spans 34 regions and 22 languages, covering 8 evaluation sub-metrics through LLM-as-a-Judge and 5 jailbreak attacks, with approximately 300k instances. Our evaluation of 9 LLMs reveals that the maximum attack success rate (ASR) reached 86.3% and the maximum harmfulness score was 3.5 out of 5. Notably, for English and U.S.-related topics, the defensive performance of typical multi-lingual LLMs was significantly lower than for other regions, highlighting substantial imbalances in safety across languages and regions. In addition, our analysis shows that coverage of fake news in existing safety datasets is limited and less well defended than major categories such as toxicity and social bias. Our dataset and code are available at https://github.com/kanekomasahiro/jail_news_bench.

Paperid: 317, https://arxiv.org/pdf/2603.01274.pdf GitHub

Abstract:
Machine learning accelerates molecular property prediction, yet state-of-the-art Large Language Models and Graph Neural Networks operate as black boxes. In drug discovery, where safety is critical, this opacity risks masking false correlations and excluding human expertise. Existing interpretability methods suffer from the effectiveness-trustworthiness trade-off: explanations may fail to reflect a model's true reasoning, degrade performance, or lack domain grounding. Concept Bottleneck Models (CBMs) offer a solution by projecting inputs to human-interpretable concepts before readout, ensuring that explanations are inherently faithful to the decision process. However, adapting CBMs to chemistry faces three challenges: the Relevance Gap (selecting task-relevant concepts from a large descriptor space), the Annotation Gap (obtaining concept supervision for molecular data), and the Capacity Gap (degrading performance due to bottleneck constraints). We introduce GlassMol, a model-agnostic CBM that addresses these gaps through automated concept curation and LLM-guided concept selection. Experiments across thirteen benchmarks demonstrate that \method generally matches or exceeds black-box baselines, suggesting that interpretability does not sacrifice performance and challenging the commonly assumed trade-off. Code is available at https://github.com/walleio/GlassMol.

Paperid: 318, https://arxiv.org/pdf/2603.01264.pdf GitHub

Abstract:
Adversarial training has emerged as a highly effective way to improve the robustness of deep neural networks (DNNs). It is typically conceptualized as a min-max optimization problem over model weights and adversarial perturbations, where the weights are optimized using gradient descent methods, such as SGD. In this paper, we propose a novel approach by treating model weights as random variables, which paves the way for enhancing adversarial training through \textbf{S}econd-Order \textbf{S}tatistics \textbf{O}ptimization (S$^2$O) over model weights. We challenge and relax a prevalent, yet often unrealistic, assumption in prior PAC-Bayesian frameworks: the statistical independence of weights. From this relaxation, we derive an improved PAC-Bayesian robust generalization bound. Our theoretical developments suggest that optimizing the second-order statistics of weights can substantially tighten this bound. We complement this theoretical insight by conducting an extensive set of experiments that demonstrate that S$^2$O not only enhances the robustness and generalization of neural networks when used in isolation, but also seamlessly augments other state-of-the-art adversarial training techniques. The code is available at https://github.com/Alexkael/S2O.

Paperid: 319, https://arxiv.org/pdf/2603.01236.pdf GitHub

Abstract:
Large Vision-Language Models (LVLMs) have adopted visual token pruning strategies to mitigate substantial computational overhead incurred by extensive visual token sequences. While prior works primarily focus on either attention-based or diversity-based pruning methods, in-depth analysis of these approaches' characteristics and limitations remains largely unexplored. In this work, we conduct thorough empirical analysis using effective rank (erank) as a measure of feature diversity and attention score entropy to investigate visual token processing mechanisms and analyze the strengths and weaknesses of each approach. Our analysis reveals two insights: (1) Our erank-based quantitative analysis shows that many diversity-oriented pruning methods preserve substantially less feature diversity than intended; moreover, analysis using the CHAIR dataset reveals that the diversity they do retain is closely tied to increased hallucination frequency compared to attention-based pruning. (2) We further observe that attention-based approaches are more effective on simple images where visual evidence is concentrated, while diversity-based methods better handle complex images with distributed features. Building on these empirical insights, we show that incorporating image-aware adjustments into existing hybrid pruning strategies consistently improves their performance. We also provide a minimal instantiation of our empirical findings through a simple adaptive pruning mechanism, which achieves strong and reliable performance across standard benchmarks as well as hallucination-specific evaluations. Our project page available at https://cvsp-lab.github.io/AgilePruner.

Paperid: 320, https://arxiv.org/pdf/2603.01229.pdf GitHub

Abstract:
Robotic manipulation policies have made rapid progress in recent years, yet most existing approaches give limited consideration to memory capabilities. Consequently, they struggle to solve tasks that require reasoning over historical observations and maintaining task-relevant information over time, which are common requirements in real-world manipulation scenarios. Although several memory-aware policies have been proposed, systematic evaluation of memory-dependent manipulation remains underexplored, and the relationship between architectural design choices and memory performance is still not well understood. To address this gap, we introduce RMBench, a simulation benchmark comprising 9 manipulation tasks that span multiple levels of memory complexity, enabling systematic evaluation of policy memory capabilities. We further propose Mem-0, a modular manipulation policy with explicit memory components designed to support controlled ablation studies. Through extensive simulation and real-world experiments, we identify memory-related limitations in existing policies and provide empirical insights into how architectural design choices influence memory performance. The website is available at https://rmbench.github.io/.

Paperid: 321, https://arxiv.org/pdf/2603.01209.pdf GitHub

Abstract:
Tool-augmented LLMs are increasingly deployed as agents that interleave natural-language reasoning with executable Python actions, as in CodeAct-style frameworks. In deployment, these agents rely on runtime state that persists across steps. By contrast, the traces used to post-train these models rarely encode how interpreter state is managed. We ask whether interpreter persistence is merely a runtime scaffold, or a property of the training data that shapes how agents learn to use the interpreter. We isolate state persistence as a training-time variable. We introduce Opaque Knapsack, a procedurally generated family of partially observable optimization tasks designed to prevent one-shot solutions. Item attributes and constraints are hidden behind budgeted tool calls, forcing multi-turn control flow and iterative state revision. Holding task instances, prompts, tools, model, and supervision fixed, we generate matched trajectories differing only in whether interpreter state persists across steps or resets after each action. We then fine-tune identical base models (Qwen3-8B) on each trace variant and evaluate all four train-runtime combinations. Our 2x2 cross-evaluation shows that interpreter persistence shapes how agents reach solutions, not whether they do: solution quality is statistically indistinguishable across conditions, but token cost and stability differ substantially. A persistent-trained model in a stateless runtime triggers missing-variable errors in roughly 80% of episodes; a stateless-trained model in a persistent runtime redundantly re-derives retained state, using roughly 3.5x more tokens. Interpreter persistence should be treated as a first-class semantic of agent traces. Aligning fine-tuning data with deployment runtimes improves efficiency and reduces brittle train-runtime mismatches.

Paperid: 322, https://arxiv.org/pdf/2603.01194.pdf GitHub

Abstract:
Human perceive the 3D world through 2D observations from limited viewpoints. While recent feed-forward generalizable 3D reconstruction models excel at recovering 3D structures from sparse images, their representations are often confined to observed regions, leaving unseen geometry un-modeled. This raises a key, fundamental challenge: Can we infer a complete 3D structure from partial 2D observations? We present RnG (Reconstruction and Generation), a novel feed-forward Transformer that unifies these two tasks by predicting an implicit, complete 3D representation. At the core of RnG, we propose a reconstruction-guided causal attention mechanism that separates reconstruction and generation at the attention level, and treats the KV-cache as an implicit 3D representation. Then, arbitrary poses can efficiently query this cache to render high-fidelity, novel-view RGBD outputs. As a result, RnG not only accurately reconstructs visible geometry but also generates plausible, coherent unseen geometry and appearance. Our method achieves state-of-the-art performance in both generalizable 3D reconstruction and novel view generation, while operating efficiently enough for real-time interactive applications. Project page: https://npucvr.github.io/RnG

Paperid: 323, https://arxiv.org/pdf/2603.01193.pdf GitHub

Abstract:
Training neural PDE solvers is often bottlenecked by expensive data generation or unstable physics-informed neural network (PINN) involving challenging optimization landscapes due to higher-order derivatives. To tackle this issue, we propose an alternative approach using Monte Carlo approaches to estimate the solution to the PDE as a stochastic process for weak supervision during training. Leveraging the Walk-on-Spheres method, we introduce a learning scheme called \emph{Walk-on-Spheres Neural Operator (WoS-NO)} which uses weak supervision from WoS to train any given neural operator. We propose to amortize the cost of Monte Carlo walks across the distribution of PDE instances using stochastic representations from the WoS algorithm to generate cheap, noisy, estimates of the PDE solution during training. This is formulated into a data-free physics-informed objective where a neural operator is trained to regress against these weak supervisions, allowing the operator to learn a generalized solution map for an entire family of PDEs. This strategy does not require expensive pre-computed datasets, avoids computing higher-order derivatives for loss functions that are memory-intensive and unstable, and demonstrates zero-shot generalization to novel PDE parameters and domains. Experiments show that for the same number of training steps, our method exhibits up to 8.75$\times$ improvement in $L_2$-error compared to standard physics-informed training schemes, up to 6.31$\times$ improvement in training speed, and reductions of up to 2.97$\times$ in GPU memory consumption. We present the code at https://github.com/neuraloperator/WoS-NO

Paperid: 324, https://arxiv.org/pdf/2603.01185.pdf GitHub

Abstract:
Fine-tuning large language models (LLMs) on custom datasets has become a standard approach for adapting these models to specific domains and applications. However, recent studies have shown that such fine-tuning can lead to significant degradation in the model's safety. Existing defense methods operate at the sample level and often suffer from an unsatisfactory trade-off between safety and utility. To address this limitation, we perform a systematic token-level diagnosis of safety degradation during fine-tuning. Based on this, we propose token-level data selection for safe LLM fine-tuning (TOSS), a novel framework that quantifies the safety risk of each token by measuring the loss difference between a safety-degraded model and a utility-oriented model. This token-level granularity enables accurate identification and removal of unsafe tokens, thereby preserving valuable task-specific information. In addition, we introduce a progressive refinement strategy, TOSS-Pro, which iteratively enhances the safety-degraded model's ability to identify unsafe tokens. Extensive experiments demonstrate that our approach robustly safeguards LLMs during fine-tuning while achieving superior downstream task performance, significantly outperforming existing sample-level defense methods. Our code is available at https://github.com/Polly-LYP/TOSS.

Paperid: 325, https://arxiv.org/pdf/2603.01169.pdf GitHub

Abstract:
The exponential growth of video content necessitates effective video summarization to efficiently extract key information from long videos. However, current approaches struggle to fully comprehend complex videos, primarily because they employ static or modality-agnostic fusion strategies. These methods fail to account for the dynamic, frame-dependent variations in modality saliency inherent in video data. To overcome these limitations, we propose TripleSumm, a novel architecture that adaptively weights and fuses the contributions of visual, text, and audio modalities at the frame level. Furthermore, a significant bottleneck for research into multimodal video summarization has been the lack of comprehensive benchmarks. Addressing this bottleneck, we introduce MoSu (Most Replayed Multimodal Video Summarization), the first large-scale benchmark that provides all three modalities. Extensive experiments demonstrate that TripleSumm achieves state-of-the-art performance, outperforming existing methods by a significant margin on four benchmarks, including MoSu. Our code and dataset are available at https://github.com/smkim37/TripleSumm.

Paperid: 326, https://arxiv.org/pdf/2603.01164.pdf GitHub

Abstract:
Image-driven video editing aims to propagate edit contents from the modified first frame to the rest frames. The existing methods usually invert the source video to noise using a pre-trained image-to-video (I2V) model and then guide the sampling process using the edited first frame. Generally, a popular choice for maintaining motion and layout from the source video is intervening in the denoising process by injecting attention during reconstruction. However, such injection often leads to unsatisfactory results, where excessive injection leads to conflicting semantics from the source video while insufficient injection brings limited source representation. Recognizing this, we propose an Editing-awaRE (REE) injection method to modulate injection intensity of each token. Specifically, we first compute the pixel difference between the source and edited first frame to form a corresponding editing mask. Next, we track the editing area throughout the entire video by using optical flow to warp the first-frame mask. Then, editing-aware feature injection intensity for each token is generated accordingly, where injection is not conducted on editing areas. Building upon REE injection, we further propose a zero-shot image-driven video editing framework with recent-emerging rectified-Flow models, dubbed FREE-Edit. Without fine-tuning or training, our FREE-Edit demonstrates effectiveness in various image-driven video editing scenarios, showing its capability to produce higher-quality outputs compared with existing techniques. Project page: https://free-edit.github.io/page/.

Paperid: 327, https://arxiv.org/pdf/2603.01161.pdf GitHub

Abstract:
Change detection (CD) in remote sensing aims to identify semantic differences between satellite images captured at different times. While deep learning has significantly advanced this field, existing approaches based on convolutional neural networks (CNNs), transformers and Selective State Space Models (SSMs) still struggle to precisely delineate change regions. In particular, traditional transformer-based methods suffer from quadratic computational complexity when applied to very high-resolution (VHR) satellite images and often perform poorly with limited training data, leading to under-utilization of the rich spatial information available in VHR imagery. We present GRAD-Former, a novel framework that enhances contextual understanding while maintaining efficiency through reduced model size. The proposed framework consists of a novel encoder with Adaptive Feature Relevance and Refinement (AFRAR) module, fusion and decoder blocks. AFRAR integrates global-local contextual awareness through two proposed components: the Selective Embedding Amplification (SEA) module and the Global-Local Feature Refinement (GLFR) module. SEA and GLFR leverage gating mechanisms and differential attention, respectively, which generates multiple softmax heaps to capture important features while minimizing the captured irreverent features. Multiple experiments across three challenging CD datasets (LEVIR-CD, CDD, DSIFN-CD) demonstrate GRAD-Former's superior performance compared to existing approaches. Notably, GRAD-Former outperforms the current state-of-the-art models across all the metrics and all the datasets while using fewer parameters. Our framework establishes a new benchmark for remote sensing change detection performance. Our code will be released at: https://github.com/Ujjwal238/GRAD-Former

Paperid: 328, https://arxiv.org/pdf/2603.01152.pdf GitHub

Abstract:
Deep-research agents are capable of executing multi-step web exploration, targeted retrieval, and sophisticated question answering. Despite their powerful capabilities, deep-research agents face two critical bottlenecks: (1) the lack of large-scale, challenging datasets with real-world difficulty, and (2) the absence of accessible, open-source frameworks for data synthesis and agent training. To bridge these gaps, we first construct DeepResearch-9K, a large-scale challenging dataset specifically designed for deep-research scenarios built from open-source multi-hop question-answering (QA) datasets via a low-cost autonomous pipeline. Notably, it consists of (1) 9000 questions spanning three difficulty levels from L1 to L3 (2) high-quality search trajectories with reasoning chains from Tongyi-DeepResearch-30B-A3B, a state-of-the-art deep-research agent, and (3) verifiable answers. Furthermore, we develop an open-source training framework DeepResearch-R1 that supports (1) multi-turn web interactions, (2) different reinforcement learning (RL) approaches, and (3) different reward models such as rule-based outcome reward and LLM-as-judge feedback. Finally, empirical results demonstrate that agents trained on DeepResearch-9K under our DeepResearch-R1 achieve state-of-the-art results on challenging deep-research benchmarks. We release the DeepResearch-9K dataset on https://huggingface.co/datasets/artillerywu/DeepResearch-9K and the code of DeepResearch-R1 on https://github.com/Applied-Machine-Learning-Lab/DeepResearch-R1.

Paperid: 329, https://arxiv.org/pdf/2603.01115.pdf GitHub

Abstract:
Foundation vision models are increasingly adopted in medical image analysis. Due to domain shift, these pretrained models misalign with medical image segmentation needs without being fully fine-tuned or lightly adapted. We introduce GuiDINO, a framework that repositions native foundation model to acting as a visual guidance generator for downstream segmentation. GuiDINO extracts visual feature representation from DINOv3 and converts them into a spatial guide mask via a lightweight TokenBook mechanism, which aggregates token-prototype similarities. This guide mask gates feature activations in multiple segmentation backbones, thereby injecting foundation-model priors while preserving the inductive biases and efficiency of medical dedicated architectures. Training relies on a guide supervision objective loss that aligns the guide mask to ground-truth regions, optionally augmented by a boundary-focused hinge loss to sharpen fine structures. GuiDINO also supports parameter-efficient adaptation through LoRA on the DINOv3 guide backbone. Across diverse medical datasets and nnUNet-style inference, GuiDINO consistently improves segmentation quality and boundary robustness, suggesting a practical alternative to fine-tuning and offering a new perspective on how foundation models can best serve medical vision. Code is available at https://github.com/Hi-FishU/GuiDINO

Paperid: 330, https://arxiv.org/pdf/2603.01108.pdf GitHub

Abstract:
Clinically reliable perception of surgical scenes is essential for advancing intelligent, context-aware intraoperative assistance such as instrument handoff guidance, collision avoidance, and workflow-aware robotic support. Existing surgical tool benchmarks primarily evaluate category-level segmentation, requiring models to detect all instances of predefined instrument classes. However, real-world clinical decisions often require resolving references to a specific instrument instance based on its functional role, spatial relation, or anatomical interaction capabilities not captured by current evaluation paradigms. We introduce GroundedSurg, the first language-conditioned, instance-level surgical grounding benchmark. Each instance pairs a surgical image with a natural-language description targeting a single instrument, accompanied by structured spatial grounding annotations including bounding boxes and point-level anchors. The dataset spans ophthalmic, laparoscopic, robotic, and open procedures, encompassing diverse instrument types, imaging conditions, and operative complexities. By jointly evaluating linguistic reference resolution and pixel-level localization, GroundedSurg enables a systematic and realistic evaluation of vision-language models in clinically realistic multi-instrument scenes. Extensive experiments demonstrate substantial performance gaps across modern segmentation and VLMs, highlighting the urgent need for clinically grounded vision-language reasoning in surgical AI systems. Code and data are publicly available at https://github.com/gaash-lab/GroundedSurg

Paperid: 331, https://arxiv.org/pdf/2603.01106.pdf GitHub

Abstract:
Reinforcement learning (RL) with group relative policy optimization (GRPO) has become a widely adopted approach for enhancing the reasoning capabilities of multimodal large language models (MLLMs). While GRPO enables long-chain reasoning without a critic, it often suffers from sparse rewards on difficult problems and advantage vanishing when group-level rewards are too consistent for overly easy or hard problems. Existing solutions (sample expansion, selective utilization, and indirect reward design) often fail to maintain enough variance in within-group reward distributions to yield clear optimization signals. To address this, we propose DIVA-GRPO, a difficulty-adaptive variant advantage method that adjusts variant difficulty distributions from a global perspective. DIVA-GRPO dynamically assesses problem difficulty, samples variants with appropriate difficulty levels, and calculates advantages across local and global groups using difficulty-weighted and normalized scaling. This alleviates reward sparsity and advantage vanishing while improving training stability. Extensive experiments on six reasoning benchmarks demonstrate that DIVA-GRPO outperforms existing approaches in training efficiency and reasoning performance. Code: https://github.com/Siaaaaaa1/DIVA-GRPO

Paperid: 332, https://arxiv.org/pdf/2603.01089.pdf GitHub

Abstract:
Large language model (LLM)-based multi-agent systems have shown strong capabilities in tasks such as code generation and collaborative reasoning. However, the effectiveness and robustness of these systems critically depend on their communication topology, which is often fixed or statically learned, ignoring real-world dynamics such as model upgrades, API (or tool) changes, or knowledge source variability. To address this limitation, we propose CARD (Conditional Agentic Graph Designer), a conditional graph-generation framework that instantiates AMACP, a protocol for adaptive multi-agent communication. CARD explicitly incorporates dynamic environmental signals into graph construction, enabling topology adaptation at both training and runtime. Through a conditional variational graph encoder and environment-aware optimization, CARD produces communication structures that are both effective and resilient to shifts in model capability or resource availability. Empirical results on HumanEval, MATH, and MMLU demonstrate that CARD consistently outperforms static and prompt-based baselines, achieving higher accuracy and robustness across diverse conditions. The source code is available at: https://github.com/Warma10032/CARD.

Paperid: 333, https://arxiv.org/pdf/2603.01083.pdf GitHub GitHub

Abstract:
Assessing the aesthetic quality of graphic design is central to visual communication, yet remains underexplored in vision language models (VLMs). We investigate whether VLMs can evaluate design aesthetics in ways comparable to humans. Prior work faces three key limitations: benchmarks restricted to narrow principles and coarse evaluation protocols, a lack of systematic VLM comparisons, and limited training data for model improvement. In this work, we introduce AesEval-Bench, a comprehensive benchmark spanning four dimensions, twelve indicators, and three fully quantifiable tasks: aesthetic judgment, region selection, and precise localization. Then, we systematically evaluate proprietary, open-source, and reasoning-augmented VLMs, revealing clear performance gaps against the nuanced demands of aesthetic assessment. Moreover, we construct a training dataset to fine-tune VLMs for this domain, leveraging human-guided VLM labeling to produce task labels at scale and indicator-grounded reasoning to tie abstract indicators to concrete design regions.Together, our work establishes the first systematic framework for aesthetic quality assessment in graphic design. Our code and dataset will be released at: \href{https://github.com/arctanxarc/AesEval-Bench}{https://github.com/arctanxarc/AesEval-Bench}

Paperid: 334, https://arxiv.org/pdf/2603.01082.pdf GitHub

Abstract:
Recent advances in multimodal large language models (MLLMs) have substantially expanded the capabilities of multimodal retrieval, enabling systems to align and retrieve information across visual and textual modalities. Yet, existing benchmarks largely focus on coarse-grained or single-condition alignment, overlooking real-world scenarios where user queries specify multiple interdependent constraints across modalities. To bridge this gap, we introduce MCMR (Multi-Conditional Multimodal Retrieval): a large-scale benchmark designed to evaluate fine-grained, multi-condition cross-modal retrieval under natural-language queries. MCMR spans five product domains: upper and bottom clothing, jewelry, shoes, and furniture. It also preserves rich long-form metadata essential for compositional matching. Each query integrates complementary visual and textual attributes, requiring models to jointly satisfy all specified conditions for relevance. We benchmark a diverse suite of MLLM-based multimodal retrievers and vision-language rerankers to assess their condition-aware reasoning abilities. Experimental results reveal: (i) distinct modality asymmetries across models; (ii) visual cues dominate early-rank precision, while textual metadata stabilizes long-tail ordering; and (iii) MLLM-based pointwise rerankers markedly improve fine-grained matching by explicitly verifying query-candidate consistency. Overall, MCMR establishes a challenging and diagnostic benchmark for advancing multimodal retrieval toward compositional, constraint-aware, and interpretable understanding. Our code and dataset is available at https://github.com/EIT-NLP/MCMR

Paperid: 335, https://arxiv.org/pdf/2603.01073.pdf GitHub

Abstract:
Diffusion-based unsupervised image registration has been explored for cardiac cine MR, but expensive multi-step inference limits practical use. We propose FlowReg, a flow-matching framework in displacement field space that achieves strong registration in as few as two steps and supports further refinement with more steps. FlowReg uses warmup-reflow training: a single-step network first acts as a teacher, then a student learns to refine from arbitrary intermediate states, removing the need for a pre-trained model as in existing methods. An Initial Guess strategy feeds back the model prediction as the next starting point, improving refinement from step two onward. On ACDC and MM2 across six tasks (including cross-dataset generalization), FlowReg outperforms the state of the art on five tasks (+0.6% mean Dice score on average), with the largest gain in the left ventricle (+1.09%), and reduces LVEF estimation error on all six tasks (-2.58 percentage points), using only 0.7% extra parameters and no segmentation labels. Code is available at https://github.com/mathpluscode/FlowReg.

Paperid: 336, https://arxiv.org/pdf/2603.01068.pdf GitHub

Abstract:
We present \textbf{LLaDA-o}, an effective and length-adaptive omni diffusion model for multimodal understanding and generation. LLaDA-o is built on a Mixture of Diffusion (MoD) framework that decouples discrete masked diffusion for text understanding and continuous diffusion for visual generation, while coupling them through a shared, simple, and efficient attention backbone that reduces redundant computation for fixed conditions. Building on MoD, we further introduce a data-centric length adaptation strategy that enables flexible-length decoding in multimodal settings without architectural changes. Extensive experiments show that LLaDA-o achieves state-of-the-art performance among omni-diffusion models on multimodal understanding and generation benchmarks, and reaches 87.04 on DPG-Bench for text-to-image generation, supporting the effectiveness of unified omni diffusion modeling. Code is available at https://github.com/ML-GSAI/LLaDA-o.

Paperid: 337, https://arxiv.org/pdf/2603.01059.pdf GitHub

Abstract:
Recent advances in large language models (LLMs) have enabled increasingly capable chatbots. However, most existing systems focus on single-user settings and do not generalize well to multi-user group chats, where agents require more proactive and accurate intervention under complex, evolving contexts. Existing approaches typically rely on LLMs for both reasoning and generation, leading to high token consumption, limited scalability, and potential privacy risks. To address these challenges, we propose GroupGPT, a token-efficient and privacy-preserving agentic framework for multi-user chat assistant. GroupGPT adopts a small-large model collaborative architecture to decouple intervention timing from response generation, enabling efficient and accurate decision-making. The framework also supports multimodal inputs, including memes, images, videos, and voice messages. We further introduce MUIR, a benchmark dataset for multi-user chat assistant intervention reasoning. MUIR contains 2,500 annotated group chat segments with intervention labels and rationales, supporting evaluation of timing accuracy and response quality. We evaluate a range of models on MUIR, from large language models to smaller counterparts. Extensive experiments demonstrate that GroupGPT produces accurate and well-timed responses, achieving an average score of 4.72/5.0 in LLM-based evaluation, and is well received by users across diverse group chat scenarios. Moreover, GroupGPT reduces token usage by up to 3 times compared to baseline methods, while providing privacy sanitization of user messages before cloud transmission. Code is available at: https://github.com/Eliot-Shen/GroupGPT .

Paperid: 338, https://arxiv.org/pdf/2603.01050.pdf GitHub

Abstract:
We aim to develop a multimodal research agent capable of explicit reasoning and planning, multi-tool invocation, and cross-modal information synthesis, enabling it to conduct deep research tasks. However, we observe three main challenges in developing such agents: (1) scarcity of search-intensive multimodal QA data, (2) lack of effective search trajectories, and (3) prohibitive cost of training with online search APIs. To tackle them, we first propose Hyper-Search, a hypergraph-based QA generation method that models and connects visual and textual nodes within and across modalities, enabling to generate search-intensive multimodal QA pairs that require invoking various search tools to solve. Second, we introduce DR-TTS, which first decomposes search-involved tasks into several categories according to search tool types, and respectively optimize specialized search tool experts for each tool. It then recomposes tool experts to jointly explore search trajectories via tree search, producing trajectories that successfully solve complex tasks using various search tools. Third, we build an offline search engine supporting multiple search tools, enabling agentic reinforcement learning without using costly online search APIs. With the three designs, we develop MM-DeepResearch, a powerful multimodal deep research agent, and extensive results shows its superiority across benchmarks. Code is available at https://github.com/HJYao00/MM-DeepResearch

Paperid: 339, https://arxiv.org/pdf/2603.01042.pdf GitHub

Abstract:
Large Language Models (LLMs) have demonstrated remarkable success in general-purpose reasoning. However, they still struggle to understand and reason about time series data, which limits their effectiveness in decision-making scenarios that depend on temporal dynamics. In this paper, we propose Thoth, the first family of mid-trained LLMs with general-purpose time series understanding capabilities. As a pivotal intermediate stage, mid-training achieves task- and domain-agnostic alignment between time series and natural language, for which we construct Book-of-Thoth, a high-quality, time-series-centric mid-training corpus. Book-of-Thoth enables both time-series-to-text and text-to-time-series generation, equipping LLMs with a foundational grasp of temporal patterns. To better evaluate advanced reasoning capabilities, we further present KnoTS, a novel benchmark of knowledge-intensive time series understanding, designed for joint reasoning over temporal patterns and domain knowledge. Extensive experiments demonstrate that mid-training with Book-of-Thoth enables Thoth to significantly outperform its base model and advanced LLMs across a range of time series question answering benchmarks. Moreover, Thoth exhibits superior capabilities when fine-tuned under data scarcity, underscoring the effectiveness of mid-training for time series understanding. Code is available at: https://github.com/thuml/Thoth.

Paperid: 340, https://arxiv.org/pdf/2603.01034.pdf GitHub

Abstract:
Tensor Ring (TR) decomposition is a powerful tool for high-order data modeling, but is inherently restricted to discrete forms defined on fixed meshgrids. In this work, we propose a TR functional decomposition for both meshgrid and non-meshgrid data, where factors are parameterized by Implicit Neural Representations (INRs). However, optimizing this continuous framework to capture fine-scale details is intrinsically difficult. Through a frequency-domain analysis, we demonstrate that the spectral structure of TR factors determines the frequency composition of the reconstructed tensor and limits the high-frequency modeling capacity. To mitigate this, we propose a reparameterized TR functional decomposition, in which each TR factor is a structured combination of a learnable latent tensor and a fixed basis. This reparameterization is theoretically shown to improve the training dynamics of TR factor learning. We further derive a principled initialization scheme for the fixed basis and prove the Lipschitz continuity of our proposed model. Extensive experiments on image inpainting, denoising, super-resolution, and point cloud recovery demonstrate that our method achieves consistently superior performance over existing approaches. Code is available at https://github.com/YangyangXu2002/RepTRFD.

Paperid: 341, https://arxiv.org/pdf/2603.01029.pdf GitHub

Abstract:
Safe autonomous systems in complex environments require robust road anomaly segmentation to identify unknown obstacles. However, existing approaches often rely on pixel-level statistics to determine whether a region appears anomalous. This reliance leads to high false-positive rates on semantically normal background regions such as sky or vegetation, and poor recall of true Out-of-distribution (OOD) instances, thereby posing safety risks for robotic perception and decision-making. To address these challenges, we propose VL-Anomaly, a vision-language anomaly segmentation framework that incorporates semantic priors from pre-trained Vision-Language Models (VLMs). Specifically, we design a prompt learning-driven alignment module that adapts Mask2Forme's visual features to CLIP text embeddings of known categories, effectively suppressing spurious anomaly responses in background regions. At inference time, we further introduce a multi-source inference strategy that integrates text-guided similarity, CLIP-based image-text similarity and detector confidence, enabling more reliable anomaly prediction by leveraging complementary information sources. Extensive experiments demonstrate that VL-Anomaly achieves state-of-the-art performance on benchmark datasets including RoadAnomaly, SMIYC and Fishyscapes.Code is released on https://github.com/NickHezhuolin/VL-aligner-Road-anomaly-segment.

Paperid: 342, https://arxiv.org/pdf/2603.01028.pdf GitHub

Abstract:
Implicit Neural Representations (INRs) have emerged as a powerful paradigm for various signal processing tasks, but their inherent spectral bias limits the ability to capture high-frequency details. Existing methods partially mitigate this issue by using Fourier-based features, which usually rely on fixed frequency bases. This forces multi-layer perceptrons (MLPs) to inefficiently compose the required frequencies, thereby constraining their representational capacity. To address this limitation, we propose Content-Aware Frequency Encoding (CAFE), which builds upon Fourier features through multiple parallel linear layers combined via a Hadamard product. CAFE can explicitly and efficiently synthesize a broader range of frequency bases, while the learned weights enable the selection of task-relevant frequencies. Furthermore, we extend this framework to CAFE+, which incorporates Chebyshev features as a complementary component to Fourier bases. This combination provides a stronger and more stable frequency representation. Extensive experiments across multiple benchmarks validate the effectiveness and efficiency of our approach, consistently achieving superior performance over existing methods. Our code is available at https://github.com/JunboKe0619/CAFE.

Paperid: 343, https://arxiv.org/pdf/2603.01013.pdf GitHub

Abstract:
In the social sciences, it is often necessary to debias studies and surveys before valid conclusions can be drawn. Debiasing algorithms enable the computational removal of bias using sample weights. However, an issue arises when only a subset of features is highly biased, while the rest is already representative. Algorithms need to strongly alter the sample distribution to manage a few highly biased features, which can in turn introduce bias into already representative variables. To address this issue, we developed a method that uses feature weights to minimize the impact of highly biased features on the computation of sample weights. Our algorithm is based on Maximum Representative Subsampling (MRS), which debiases datasets by aligning a non-representative sample with a representative one through iterative removal of elements to create a representative subsample. The new algorithm, named feature-weighted MRS (FW-MRS), decreases the emphasis on highly biased features, allowing it to retain more instances for downstream tasks. The feature weights are derived from the feature importance of a domain classifier trained to differentiate between the representative and non-representative datasets. We validated FW-MRS using eight tabular datasets, each of which we artificially biased. Biased features can be important for downstream tasks, and focusing less on them could lead to a decline in generalization. For this reason, we assessed the generalization performance of FW-MRS on downstream tasks and found no statistically significant differences. Additionally, FW-MRS was applied to a real-world dataset from the social sciences. The source code is available at https://github.com/kramerlab/FeatureWeightDebiasing.

Paperid: 344, https://arxiv.org/pdf/2603.01012.pdf GitHub

Abstract:
Repository-scale code reasoning is a cornerstone of modern AI-assisted software engineering, enabling Large Language Models (LLMs) to handle complex workflows from program comprehension to complex debugging. However, balancing accuracy with context cost remains a significant bottleneck, as existing agentic approaches often waste computational resources through inefficient, iterative full-text exploration. To address this, we introduce FastCode, a framework that decouples repository exploration from content consumption. FastCode utilizes a structural scouting mechanism to navigate a lightweight semantic-structural map of the codebase, allowing the system to trace dependencies and pinpoint relevant targets without the overhead of full-text ingestion. By leveraging structure-aware navigation tools regulated by a cost-aware policy, the framework constructs high-value contexts in a single, optimized step. Extensive evaluations on the SWE-QA, LongCodeQA, LOC-BENCH, and GitTaskBench benchmarks demonstrate that FastCode consistently outperforms state-of-the-art baselines in reasoning accuracy while significantly reducing token consumption, validating the efficiency of scouting-first strategies for large-scale code reasoning. Source code is available at https://github.com/HKUDS/FastCode.

Paperid: 345, https://arxiv.org/pdf/2603.01000.pdf GitHub

Abstract:
Motion transfer has emerged as a promising direction for controllable video generation, yet existing methods largely focus on single-object scenarios and struggle when multiple objects require distinct motion patterns. In this work, we present FlexiMMT, the first implicit image-to-video (I2V) motion transfer framework that explicitly enables multi-object, multi-motion transfer. Given a static multi-object image and multiple reference videos, FlexiMMT independently extracts motion representations and accurately assigns them to different objects, supporting flexible recombination and arbitrary motion-to-object mappings. To address the core challenge of cross-object motion entanglement, we introduce a Motion Decoupled Mask Attention Mechanism that uses object-specific masks to constrain attention, ensuring that motion and text tokens only influence their designated regions. We further propose a Differentiated Mask Propagation Mechanism that derives object-specific masks directly from diffusion attention and progressively propagates them across frames efficiently. Extensive experiments demonstrate that FlexiMMT achieves precise, compositional, and state-of-the-art performance in I2V-based multi-object multi-motion transfer.

Paperid: 346, https://arxiv.org/pdf/2603.00960.pdf GitHub

Abstract:
Modern web applications are increasingly produced through AI-assisted development and rapid no-code deployment pipelines, widening the gap between accelerating software velocity and the limited adaptability of existing security tooling. Pattern-driven scanners fail to reason about novel contexts, while emerging LLM-based penetration testers rely on unconstrained exploration, yielding high cost, unstable behavior, and poor reproducibility. We introduce AWE, a memory-augmented multi-agent framework for autonomous web penetration testing that embeds structured, vulnerability-specific analysis pipelines within a lightweight LLM orchestration layer. Unlike general-purpose agents, AWE couples context aware payload mutations and generations with persistent memory and browser-backed verification to produce deterministic, exploitation-driven results. Evaluated on the 104-challenge XBOW benchmark, AWE achieves substantial gains on injection-class vulnerabilities - 87% XSS success (+30.5% over MAPTA) and 66.7% blind SQL injection success (+33.3%) - while being much faster, cheaper, and more token-efficient than MAPTA, despite using a midtier model (Claude Sonnet 4) versus MAPTA's GPT-5. MAPTA retains higher overall coverage due to broader exploratory capabilities, underscoring the complementary strengths of specialized and general-purpose architectures. Our results demonstrate that architecture matters as much as model reasoning capabilities: integrating LLMs into principled, vulnerability-aware pipelines yields substantial gains in accuracy, efficiency, and determinism for injection-class exploits. The source code for AWE is available at: https://github.com/stuxlabs/AWE

Paperid: 347, https://arxiv.org/pdf/2603.00958.pdf GitHub

Abstract:
With recent advances in Text-to-Speech (TTS) systems, synthetic audiobook narration has seen increased interest, reaching unprecedented levels of naturalness. However, larger gaps remain in synthetic narration systems' ability to impersonate fictional characters, and convey complex emotions or prosody. A promising direction to enhance character identification is the assignment of plausible voices to each fictional characters in a book. This step typically requires complex inference of attributes in book-length contexts, such as a character's age, gender, origin or physical health, which in turns requires dedicated benchmark datasets to evaluate extraction systems' performances. We present S-VoCAL (Speaking Voice Character Attributes in Literature), the first dataset and evaluation framework dedicated to evaluate the inference of voice-related fictional character attributes. S-VoCAL entails 8 attributes grounded in sociophonetic studies, and 952 character-book pairs derived from Project Gutenberg. Its evaluation framework addresses the particularities of each attribute, and includes a novel similarity metric based on recent Large Language Models embeddings. We demonstrate the applicability of S-VoCAL by applying a simple Retrieval-Augmented Generation (RAG) pipeline to the task of inferring character attributes. Our results suggest that the RAG pipeline reliably infers attributes such as Age or Gender, but struggles on others such as Origin or Physical Health. The dataset and evaluation code are available at https://github.com/AbigailBerthe/S-VoCAL .

Paperid: 348, https://arxiv.org/pdf/2603.00949.pdf GitHub

Abstract:
Recently, Instant Neural Graphics Primitives (Instant-NGP) has achieved significant success in rapid 3D scene reconstruction, but securely embedding high-capacity hidden data, such as an entire 3D scene, remains a challenge. Existing methods rely on external decoders, require architectural modifications, and suffer from limited capacity, which makes them easily detectable. We propose a novel parameter-free 3D Cryptographic Steganography using Instant-NGP (StegoNGP), which leverages the Instant-NGP hash encoding function as a key-controlled scene switcher. By associating a default key with a cover scene and a secret key with a hidden scene, our method trains a single model to interweave both representations within the same network weights. The resulting model is indistinguishable from a standard Instant-NGP in architecture and parameter count. We also introduce an enhanced Multi-Key scheme, which assigns multiple independent keys across hash levels, dramatically expanding the key space and providing high robustness against partial key disclosure attacks. Experimental results demonstrated that StegoNGP can hide a complete high-quality 3D scene with strong imperceptibility and security, providing a new paradigm for high-capacity, undetectable information hiding in neural fields. The code can be found at https://github.com/jiang-wenxiang/StegoNGP.

Paperid: 349, https://arxiv.org/pdf/2603.00947.pdf GitHub GitHub

Abstract:
Virtual try-on (VTON) has recently achieved impressive visual fidelity, but most existing systems require uploading personal photos to cloud-based GPUs, raising privacy concerns and limiting on-device deployment. To address this, we present Mobile-VTON, a high-quality, privacy-preserving framework that enables fully offline virtual try-on on commodity mobile devices using only a single user image and a garment image. Mobile-VTON introduces a modular TeacherNet-GarmentNet-TryonNet (TGT) architecture that integrates knowledge distillation, garment-conditioned generation, and garment alignment into a unified pipeline optimized for on-device efficiency. Within this framework, we propose a Feature-Guided Adversarial (FGA) Distillation strategy that combines teacher supervision with adversarial learning to better match real-world image distributions. GarmentNet is trained with a trajectory-consistency loss to preserve garment semantics across diffusion steps, while TryonNet uses latent concatenation and lightweight cross-modal conditioning to enable robust garment-to-person alignment without large-scale pretraining. By combining these components, Mobile-VTON achieves high-fidelity generation with low computational overhead. Experiments on VITON-HD and DressCode at 1024 x 768 show that it matches or outperforms strong server-based baselines while running entirely offline. These results demonstrate that high-quality VTON is not only feasible but also practical on-device, offering a secure solution for real-world applications. Code and project page are available at https://zhenchenwan.github.io/Mobile-VTON/.

Paperid: 350, https://arxiv.org/pdf/2603.00919.pdf GitHub

Abstract:
Large language models (LLMs) have shown great promise for autonomous driving. However, discretizing numbers into tokens limits precise numerical reasoning, fails to reflect the positional significance of digits in the training objective, and makes it difficult to achieve both decoding efficiency and numerical precision. These limitations affect both the processing of sensor measurements and the generation of precise control commands, creating a fundamental barrier for deploying LLM-based autonomous driving systems. In this paper, we introduce DriveCode, a novel numerical encoding method that represents numbers as dedicated embeddings rather than discrete text tokens. DriveCode employs a number projector to map numbers into the language model's hidden space, enabling seamless integration with visual and textual features in a unified multimodal sequence. Evaluated on OmniDrive, DriveGPT4, and DriveGPT4-V2 datasets, DriveCode demonstrates superior performance in trajectory prediction and control signal generation, confirming its effectiveness for LLM-based autonomous driving systems.

Paperid: 351, https://arxiv.org/pdf/2603.00918.pdf GitHub

Abstract:
Text-to-image generation powers content creation across design, media, and data augmentation. Post-training of text-to-image generative models is a promising path to better match human preferences, factuality, and improved aesthetics. We introduce SOLACE (Adaptive Rewarding by self-Confidence), a post-training framework that replaces external reward supervision with an internal self-confidence signal, obtained by evaluating how accurately the model recovers injected noise under self-denoising probes. SOLACE converts this intrinsic signal into scalar rewards, enabling fully unsupervised optimization without additional datasets, annotators, or reward models. Empirically, by reinforcing high-confidence generations, SOLACE delivers consistent gains in compositional generation, text rendering and text-image alignment over the baseline. We also find that integrating SOLACE with external rewards results in a complementary improvement, with alleviated reward hacking.

Paperid: 352, https://arxiv.org/pdf/2603.00912.pdf GitHub

Abstract:
Current multi-view indoor 3D object detectors rely on sensor geometry that is costly to obtain (i.e., precisely calibrated multi-view camera poses) to fuse multi-view information into a global scene representation, limiting deployment in real-world scenes. We target a more practical setting: Sensor-Geometry-Free (SG-Free) multi-view indoor 3D object detection, where there are no sensor-provided geometric inputs (multi-view poses or depth). Recent Visual Geometry Grounded Transformer (VGGT) shows that strong 3D cues can be inferred directly from images. Building on this insight, we present VGGT-Det, the first framework tailored for SG-Free multi-view indoor 3D object detection. Rather than merely consuming VGGT predictions, our method integrates VGGT encoder into a transformer-based pipeline. To effectively leverage both the semantic and geometric priors from inside VGGT, we introduce two novel key components: (i) Attention-Guided Query Generation (AG): exploits VGGT attention maps as semantic priors to initialize object queries, improving localization by focusing on object regions while preserving global spatial structure; (ii) Query-Driven Feature Aggregation (QD): a learnable See-Query interacts with object queries to 'see' what they need, and then dynamically aggregates multi-level geometric features across VGGT layers that progressively lift 2D features into 3D. Experiments show that VGGT-Det significantly surpasses the best-performing method in the SG-Free setting by 4.4 and 8.6 mAP@0.25 on ScanNet and ARKitScenes, respectively. Ablation study shows that VGGT's internally learned semantic and geometric priors can be effectively leveraged by our AG and QD.

Paperid: 353, https://arxiv.org/pdf/2603.00908.pdf GitHub

Abstract:
Underwater optical imaging is severely hindered by scattering, but polarization imaging offers the unique dual advantages of descattering and shape-from-polarization (SfP) 3D reconstruction. To exploit these advantages, this paper proposes UD-SfPNet, an underwater descattering shape-from-polarization network that leverages polarization cues for improved 3D surface normal prediction. The framework jointly models polarization-based image descattering and SfP normal estimation in a unified pipeline, avoiding error accumulation from sequential processing and enabling global optimization across both tasks. UD-SfPNet further incorporates a novel color embedding module to enhance geometric consistency by exploiting the relationship between color encodings and surface orientation. A detail enhancement convolution module is also included to better preserve high-frequency geometric details that are lost under scattering. Experiments on the MuS-Polar3D dataset show that the proposed method significantly improves reconstruction accuracy, achieving a mean surface normal angular error of 15.12$^\circ$ (the lowest among compared methods). These results confirm the efficacy of combining descattering with polarization-based shape inference, and highlight the practical significance and potential applications of UD-SfPNet for optical 3D imaging in challenging underwater environments. The code is available at https://github.com/WangPuyun/UD-SfPNet.

Paperid: 354, https://arxiv.org/pdf/2603.00906.pdf GitHub

Abstract:
Look-Up Table based methods have emerged as a promising direction for efficient image restoration tasks. Recent LUT-based methods focus on improving their performance by expanding the receptive field. However, they inevitably introduce extra computational and storage overhead, which hinders their deployment in edge devices. To address this issue, we propose ShiftLUT, a novel framework that attains the largest receptive field among all LUT-based methods while maintaining high efficiency. Our key insight lies in three complementary components. First, Learnable Spatial Shift module (LSS) is introduced to expand the receptive field by applying learnable, channel-wise spatial offsets on feature maps. Second, we propose an asymmetric dual-branch architecture that allocates more computation to the information-dense branch, substantially reducing inference latency without compromising restoration quality. Finally, we incorporate a feature-level LUT compression strategy called Error-bounded Adaptive Sampling (EAS) to minimize the storage overhead. Compared to the previous state-of-the-art method TinyLUT, ShiftLUT achieves a 3.8$\times$ larger receptive field and improves an average PSNR by over 0.21 dB across multiple standard benchmarks, while maintaining a small storage size and inference time. The code is available at: https://github.com/Sailor-t/ShiftLUT .

Paperid: 355, https://arxiv.org/pdf/2603.00903.pdf GitHub

Abstract:
Inspired by the human learning and memory system, particularly the interplay between the hippocampus and cerebral cortex, this study proposes a dual-learner framework comprising a fast learner and a meta learner to address continual Reinforcement Learning~(RL) problems. These two learners are coupled to perform distinct yet complementary roles: the fast learner focuses on knowledge transfer, while the meta learner ensures knowledge integration. In contrast to traditional multi-task RL approaches that share knowledge through average return maximization, our meta learner incrementally integrates new experiences by explicitly minimizing catastrophic forgetting, thereby supporting efficient cumulative knowledge transfer for the fast learner. To facilitate rapid adaptation in new environments, we introduce an adaptive meta warm-up mechanism that selectively harnesses past knowledge. We conduct experiments in various pixel-based and continuous control benchmarks, revealing the superior performance of continual learning for our proposed dual-learner approach relative to baseline methods. The code is released in https://github.com/datake/FAME.

Paperid: 356, https://arxiv.org/pdf/2603.00897.pdf GitHub

Abstract:
Large language models are increasingly used to produce runnable software. In practice, security is often addressed through a Detect--Repair--Verify (DRV) loop that detects issues, applies fixes, and verifies the result. This work studies such a workflow for project-level artifacts and addresses four gaps: L1, the lack of project-level benchmarks with executable function and security tests; L2, limited evidence on pipeline-level effectiveness beyond studying detection or repair alone; L3, unclear reliability of detection reports as repair guidance; and L4, uncertain repair trustworthiness and side effects under verification. A new benchmark dataset\footnote{https://github.com/Hahappyppy2024/EmpricalVDR} is introduced, consisting of runnable web-application projects paired with functional tests and targeted security tests, and supporting three prompt granularities at the project, requirement, and function level. The evaluation compares generation-only, single-pass DRV, and bounded iterative DRV variants under comparable budget constraints. Outcomes are measured by secure and correct yield using test-grounded verification, and intermediate artifacts are analyzed to assess report actionability and post-repair failure modes such as regressions, semantic drift, and newly introduced security issues.

Paperid: 357, https://arxiv.org/pdf/2603.00887.pdf GitHub

Abstract:
Volume Electron Microscopy (VEM) is crucial for 3D tissue imaging but often produces anisotropic data with poor axial resolution, hindering visualization and downstream analysis. Existing methods for isotropic reconstruction often suffer from neglecting abundant axial information and employing simple downsampling to simulate anisotropic data. To address these limitations, we propose VEMamba, an efficient framework for isotropic reconstruction. The core of VEMamba is a novel 3D Dependency Reordering paradigm, implemented via two key components: an Axial-Lateral Chunking Selective Scan Module (ALCSSM), which intelligently re-maps complex 3D spatial dependencies (both axial and lateral) into optimized 1D sequences for efficient Mamba-based modeling, explicitly enforcing axial-lateral consistency; and a Dynamic Weights Aggregation Module (DWAM) to adaptively aggregate these reordered sequence outputs for enhanced representational power. Furthermore, we introduce a realistic degradation simulation and then leverage Momentum Contrast (MoCo) to integrate this degradation-aware knowledge into the network for superior reconstruction. Extensive experiments on both simulated and real-world anisotropic VEM datasets demonstrate that VEMamba achieves highly competitive performance across various metrics while maintaining a lower computational footprint. The source code is available on GitHub: https://github.com/I2-Multimedia-Lab/VEMamba

Paperid: 358, https://arxiv.org/pdf/2603.00881.pdf GitHub

Abstract:
Segmentation of the main coronary artery from X-ray coronary angiography (XCA) sequences is crucial for the diagnosis of coronary artery diseases. However, this task is challenging due to issues such as blurred boundaries, inconsistent radiation contrast, complex motion patterns, and a lack of annotated images for training. Although Semi-Supervised Learning (SSL) can alleviate the annotation burden, conventional methods struggle with complicated temporal dynamics and unreliable uncertainty quantification. To address these challenges, we propose SAM3-based Teacher-student framework with Motion-Aware consistency and Progressive Confidence Regularization (SMART), a semi-supervised vessel segmentation approach for X-ray angiography videos. First, our method utilizes SAM3's unique promptable concept segmentation design and innovates a SAM3-based teacher-student framework to maximize the performance potential of both the teacher and the student. Second, we enhance segmentation by integrating the vessel mask warping technique and motion consistency loss to model complex vessel dynamics. To address the issue of unreliable teacher predictions caused by blurred boundaries and minimal contrast, we further propose a progressive confidence-aware consistency regularization to mitigate the risk of unreliable outputs. Extensive experiments on three datasets of XCA sequences from different institutions demonstrate that SMART achieves state-of-the-art performance while requiring significantly fewer annotations, making it particularly valuable for real-world clinical applications where labeled data is scarce. Our code is available at: https://github.com/qimingfan10/SMART.

Paperid: 359, https://arxiv.org/pdf/2603.00876.pdf GitHub GitHub

Abstract:
Large language models (LLMs) have demonstrated significant reasoning capabilities in scientific discovery but struggle to bridge the gap to physical execution in wet-labs. In these irreversible environments, probabilistic hallucinations are not merely incorrect, but also cause equipment damage or experimental failure. To address this, we propose \textbf{BioProAgent}, a neuro-symbolic framework that anchors probabilistic planning in a deterministic Finite State Machine (FSM). We introduce a State-Augmented Planning mechanism that enforces a rigorous \textit{Design-Verify-Rectify} workflow, ensuring hardware compliance before execution. Furthermore, we address the context bottleneck inherent in complex device schemas by \textit{Semantic Symbol Grounding}, reducing token consumption by $\sim$6$\times$ through symbolic abstraction. In the extended BioProBench benchmark, BioProAgent achieves 95.6\% physical compliance (compared to 21.0\% for ReAct), demonstrating that neuro-symbolic constraints are essential for reliable autonomy in irreversible physical environments. \footnote{Code at https://github.com/YuyangSunshine/bioproagent and project at https://yuyangsunshine.github.io/BioPro-Project/}

Paperid: 360, https://arxiv.org/pdf/2603.00853.pdf GitHub GitHub

Abstract:
We propose a simple yet effective UHDPromer, a neural discrimination-prompted Transformer, for Ultra-High-Definition (UHD) image restoration and enhancement. Our UHDPromer is inspired by an interesting observation that there implicitly exist neural differences between high-resolution and low-resolution features, and exploring such differences can facilitate low-resolution feature representation. To this end, we first introduce Neural Discrimination Priors (NDP) to measure the differences and then integrate NDP into the proposed Neural Discrimination-Prompted Attention (NDPA) and Neural Discrimination-Prompted Network (NDPN). The proposed NDPA re-formulates the attention by incorporating NDP to globally perceive useful discrimination information, while the NDPN explores a continuous gating mechanism guided by NDP to selectively permit the passage of beneficial content. To enhance the quality of restored images, we propose a super-resolution-guided reconstruction approach, which is guided by super-resolving low-resolution features to facilitate final UHD image restoration. Experiments show that UHDPromer achieves the best computational efficiency while still maintaining state-of-the-art performance on $3$ UHD image restoration and enhancement tasks, including low-light image enhancement, image dehazing, and image deblurring. The source codes and pre-trained models will be made available at https://github.com/supersupercong/uhdpromer.

Paperid: 361, https://arxiv.org/pdf/2603.00840.pdf GitHub

Abstract:
Nested named entity recognition identifies entities contained within other entities, but requires expensive multi-level annotation. While flat NER corpora exist abundantly, nested resources remain scarce. We investigate whether models can learn nested structure from flat annotations alone, evaluating four approaches: string inclusions (substring matching), entity corruption (pseudo-nested data), flat neutralization (reducing false negative signal), and a hybrid fine-tuned + LLM pipeline. On NEREL, a Russian benchmark with 29 entity types where 21% of entities are nested, our best combined method achieves 26.37% inner F1, closing 40% of the gap to full nested supervision. Code is available at https://github.com/fulstock/Learning-from-Flat-Annotations.

Paperid: 362, https://arxiv.org/pdf/2603.00828.pdf GitHub

Abstract:
In recent years, various methods have been proposed for mesh analysis, each offering distinct advantages and often excelling on different object classes. We present a novel Mixture of Experts (MoE) framework designed to harness the complementary strengths of these diverse approaches. We propose a new gate architecture that encourages each expert to specialise in the classes it excels in. Our design is guided by two key ideas: (1) random walks over the mesh surface effectively capture the regions that individual experts attend to, and (2) an attention mechanism that enables the gate to focus on the areas most informative for each expert's decision-making. To further enhance performance, we introduce a dynamic loss balancing scheme that adjusts a trade-off between diversity and similarity losses throughout the training, where diversity prompts expert specialization, and similarity enables knowledge sharing among the experts. Our framework achieves state-of-the-art results in mesh classification, retrieval, and semantic segmentation tasks. Our code is available at: https://github.com/amirbelder/MME-Mixture-of-Mesh-Experts.

Paperid: 363, https://arxiv.org/pdf/2603.00812.pdf GitHub

Abstract:
Work introduces a hierarchical binary tree-based reduction that replaces standard self-attention. The core idea is to use a recursive Gated Linear Unit merge operation, achieving O(n) total merge operations O(log n) parallel depth O(n d^2) total work and O(n) space complexity. In these experiments, the model significantly outperforms standard Transformers in both convergence speed and accuracy on long-range structural dependencies, specifically where hierarchical inductive bias is critical.

Paperid: 364, https://arxiv.org/pdf/2603.00805.pdf GitHub

Abstract:
The proliferation of neural radiance field (NeRF) research requires significant efforts to reimplement papers before building upon them. We introduce NERFIFY, a multi-agent framework that reliably converts NeRF research papers into trainable Nerfstudio plugins, in contrast to generic paper-to-code methods and frontier models like GPT-5 that usually fail to produce runnable code. NERFIFY achieves domain-specific executability through six key innovations: (1) Context-free grammar (CFG): LLM synthesis is constrained by Nerfstudio formalized as a CFG, ensuring generated code satisfies architectural invariants. (2) Graph-of-Thought code synthesis: Specialized multi-file-agents generate repositories in topological dependency order, validating contracts and errors at each node. (3) Compositional citation recovery: Agents automatically retrieve and integrate components (samplers, encoders, proposal networks) from citation graphs of references. (4) Visual feedback: Artifacts are diagnosed through PSNR-minima ROI analysis, cross-view geometric validation, and VLM-guided patching to iteratively improve quality. (5) Knowledge enhancement: Beyond reproduction, methods can be improved with novel optimizations. (6) Benchmarking: An evaluation framework is designed for NeRF paper-to-code synthesis across 30 diverse papers. On papers without public implementations, NERFIFY achieves visual quality matching expert human code (+/-0.5 dB PSNR, +/-0.2 SSIM) while reducing implementation time from weeks to minutes. NERFIFY demonstrates that a domain-aware design enables code translation for complex vision papers, potentiating accelerated and democratized reproducible research. Code, data and implementations will be publicly released.

Paperid: 365, https://arxiv.org/pdf/2603.00792.pdf GitHub GitHub

Abstract:
Fluid-solid interaction (FSI) problems are fundamental in many scientific and engineering applications, yet effectively capturing the highly nonlinear two-way interactions remains a significant challenge. Most existing deep learning methods are limited to simplified one-way FSI scenarios, often assuming rigid and static solid to reduce complexity. Even in two-way setups, prevailing approaches struggle to capture dynamic, heterogeneous interactions due to the lack of cross-domain awareness. In this paper, we introduce \textbf{Fisale}, a data-driven framework for handling complex two-way \textbf{FSI} problems. It is inspired by classical numerical methods, namely the Arbitrary Lagrangian-Eulerian (\textbf{ALE}) method and the partitioned coupling algorithm. Fisale explicitly models the coupling interface as a distinct component and leverages multiscale latent ALE grids to provide unified, geometry-aware embeddings across domains. A partitioned coupling module (PCM) further decomposes the problem into structured substeps, enabling progressive modeling of nonlinear interdependencies. Compared to existing models, Fisale introduces a more flexible framework that iteratively handles complex dynamics of solid, fluid and their coupling interface on a unified representation, and enables scalable learning of complex two-way FSI behaviors. Experimentally, Fisale excels in three reality-related challenging FSI scenarios, covering 2D, 3D and various tasks. The code is available at \href{https://github.com/therontau0054/Fisale}.

Paperid: 366, https://arxiv.org/pdf/2603.00732.pdf GitHub GitHub

Abstract:
Planning physically feasible dexterous hand manipulation is a central challenge in robotic manipulation and Embodied AI. Prior work typically relies on object-centric cues or precise hand-object interaction sequences, foregoing the rich, compositional guidance of open-vocabulary instruction. We introduce UniHM, the first framework for unified dexterous hand manipulation guided by free-form language commands. We propose a Unified Hand-Dexterous Tokenizer that maps heterogeneous dexterous-hand morphologies into a single shared codebook, improving cross-dexterous hand generalization and scalability to new morphologies. Our vision language action model is trained solely on human-object interaction data, eliminating the need for massive real-world teleoperation datasets, and demonstrates strong generalizability in producing human-like manipulation sequences from open-ended language instructions. To ensure physical realism, we introduce a physics-guided dynamic refinement module that performs segment-wise joint optimization under generative and temporal priors, yielding smooth and physically feasible manipulation sequences. Across multiple datasets and real-world evaluations, UniHM attains state-of-the-art results on both seen and unseen objects and trajectories, demonstrating strong generalization and high physical feasibility. Our project page at \href{https://unihm.github.io/}{https://unihm.github.io/}.

Paperid: 367, https://arxiv.org/pdf/2603.00724.pdf GitHub

Abstract:
Large language model alignment via reinforcement learning depends critically on reward function quality. However, static, domain-specific reward models are often costly to train and exhibit poor generalization in out-of-distribution scenarios encountered during RL iterations. We present RLAR (Reinforcement Learning from Agent Rewards), an agent-driven framework that dynamically assigns tailored reward functions to individual queries. Specifically, RLAR transforms reward acquisition into a dynamic tool synthesis and invocation task. It leverages LLM agents to autonomously retrieve optimal reward models from the Internet and synthesize programmatic verifiers through code generation. This allows the reward system to self-evolve with the shifting data distributions during training. Experimental results demonstrate that RLAR yields consistent performance gains ranging from 10 to 60 across mathematics, coding, translation, and dialogue tasks. On RewardBench-V2, RLAR significantly outperforms static baselines and approaches the performance upper bound, demonstrating superior generalization through dynamic reward orchestration. The data and code are available on this link: https://github.com/ZhuoerFeng/RLAR.

Paperid: 368, https://arxiv.org/pdf/2603.00720.pdf GitHub

Abstract:
Fine-tuning Multimodal Large Language Models (MLLMs) with parameter-efficient methods like Low-Rank Adaptation (LoRA) is crucial for task adaptation. However, imbalanced training dynamics across modalities often lead to suboptimal accuracy due to negative interference, a challenge typically addressed with inefficient heuristic methods such as manually tuning separate learning rates. To overcome this, we introduce MARS (Multimodal Adaptive Rank Search), an approach to discover optimal rank pairs that balance training dynamics while maximizing performance. Our key innovation, a proposed framework of dual scaling laws, enables this search: one law models module-specific convergence time to prune the search space to candidates with aligned dynamics, while the other predicts final task performance to select the optimal pair from the pruned set. By re-purposing the LoRA rank as a controller for modality-specific convergence speed, MARS outperforms baseline methods and provides a robust, automated strategy for optimizing MLLM fine-tuning.

Paperid: 369, https://arxiv.org/pdf/2603.00718.pdf GitHub GitHub

Abstract:
Real-world tool-using agents operate over long-horizon workflows with recurring structure and diverse demands, where effective behavior requires not only invoking atomic tools but also abstracting, and reusing higher-level tool compositions. However, existing benchmarks mainly measure instance-level success under static tool sets, offering limited insight into agents' ability to acquire such reusable skills. We address this gap by introducing SkillCraft, a benchmark explicitly stress-test agent ability to form and reuse higher-level tool compositions, where we call Skills. SkillCraft features realistic, highly compositional tool-use scenarios with difficulty scaled along both quantitative and structural dimensions, designed to elicit skill abstraction and cross-task reuse. We further propose a lightweight evaluation protocol that enables agents to auto-compose atomic tools into executable Skills, cache and reuse them inside and across tasks, thereby improving efficiency while accumulating a persistent library of reusable skills. Evaluating state-of-the-art agents on SkillCraft, we observe substantial efficiency gains, with token usage reduced by up to 80% by skill saving and reuse. Moreover, success rate strongly correlates with tool composition ability at test time, underscoring compositional skill acquisition as a core capability.

Paperid: 370, https://arxiv.org/pdf/2603.00717.pdf GitHub

Abstract:
With the rapid advancement of generative models, generated image detection has become an important task in visual forensics. Although existing methods have achieved remarkable progress, they often rely, after training, on only a small subset of highly salient forgery cues, which limits their ability to generalize to unseen generative mechanisms. We argue that reliably generated image detection should not depend on a single decision path but should preserve multiple judgment perspectives, enabling the model to understand the differences between real and generated images from diverse viewpoints. Based on this idea, we propose an anti-feature-collapse learning framework that filters task-irrelevant components and suppresses excessive overlap among different forgery cues in the representation space, preventing discriminative information from collapsing into a few dominant feature directions. This design maintains diverse and complementary evidence within the model, reduces reliance on a small set of salient cues, and enhances robustness under unseen generative settings. Extensive experiments on multiple public benchmarks demonstrate that the proposed method significantly outperforms the state-of-the-art approaches in cross-model scenarios, achieving an accuracy improvement of 5.02% and exhibiting superior generalization and detection reliability. The source code is available at https://github.com/Yanmou-Hui/DoU.

Paperid: 371, https://arxiv.org/pdf/2603.00697.pdf GitHub

Abstract:
We present TokenSplat, a feed-forward framework for joint 3D Gaussian reconstruction and camera pose estimation from unposed multi-view images. At its core, TokenSplat introduces a Token-aligned Gaussian Prediction module that aligns semantically corresponding information across views directly in the feature space. Guided by coarse token positions and fusion confidence, it aggregates multi-scale contextual features to enable long-range cross-view reasoning and reduce redundancy from overlapping Gaussians. To further enhance pose robustness and disentangle viewpoint cues from scene semantics, TokenSplat employs learnable camera tokens and an Asymmetric Dual-Flow Decoder (ADF-Decoder) that enforces directionally constrained communication between camera and image tokens. This maintains clean factorization within a feed-forward architecture, enabling coherent reconstruction and stable pose estimation without iterative refinement. Extensive experiments demonstrate that TokenSplat achieves higher reconstruction fidelity and novel-view synthesis quality in pose-free settings, and significantly improves pose estimation accuracy compared to prior pose-free methods. Project page: https://kidleyh.github.io/tokensplat/.

Paperid: 372, https://arxiv.org/pdf/2603.00694.pdf GitHub

Abstract:
Explainability and transparent decision-making are essential for the safe deployment of autonomous driving systems. Scene captioning summarizes environmental conditions and risk factors in natural language, improving transparency, safety, and human--robot interaction. However, most existing approaches target structured urban scenarios; in off-road environments, they are vulnerable to single-modality degradations caused by rain, fog, snow, and darkness, and they lack a unified framework that jointly models structured scene captioning and path planning. To bridge this gap, we propose Wild-Drive, an efficient framework for off-road scene captioning and path planning. Wild-Drive adopts modern multimodal encoders and introduces a task-conditioned modality-routing bridge, MoRo-Former, to adaptively aggregate reliable information under degraded sensing. It then integrates an efficient large language model (LLM), together with a planning token and a gate recurrent unit (GRU) decoder, to generate structured captions and predict future trajectories. We also build the OR-C2P Benchmark, which covers structured off-road scene captioning and path planning under diverse sensor corruption conditions. Experiments on OR-C2P dataset and a self-collected dataset show that Wild-Drive outperforms prior LLM-based methods and remains more stable under degraded sensing. The code and benchmark will be publicly available at https://github.com/wangzihanggg/Wild-Drive.

Paperid: 373, https://arxiv.org/pdf/2603.00687.pdf GitHub

Abstract:
Noise and artifacts during computed tomography (CT) scans are a fundamental challenge affecting disease diagnosis. However, current methods either involve excessively long reconstruction times or rely on data-driven models for optimization, failing to adequately consider the valuable information inherent in the data itself, especially medical 3D data. This work proposes a reconstruction method under ultra-low raw data conditions, requiring no external data and avoiding lengthy pre-training processes. By leveraging spatial nonlocal similarity and the conjugate properties of the projection domain to generate pseudo-3D data for self-supervised training, high-fidelity results can be achieved in a very short time. Extensive experiments demonstrate that this method not only mitigates detector-induced ring artifacts but also exhibits unprecedented capabilities in detail recovery. This method provides a new paradigm for research using unlabeled raw projection data. Code is available at https://github.com/yqx7150/SCOUT.

Paperid: 374, https://arxiv.org/pdf/2603.00686.pdf GitHub

Abstract:
Large Language Models have evolved from single-round generators into long-horizon agents, capable of complex text synthesis scenarios. However, current evaluation frameworks lack the ability to assess the actual synthesis operations, such as outlining, drafting, and editing. Consequently, they fail to evaluate the actual and detailed capabilities of LLMs. To bridge this gap, we introduce RAVEL, an agentic framework that enables the LLM testers to autonomously plan and execute typical synthesis operations, including outlining, drafting, reviewing, and refining. Complementing this framework, we present C3EBench, a comprehensive benchmark comprising 1,258 samples derived from professional human writings. We utilize a "reverse-engineering" pipeline to isolate specific capabilities across four tasks: Cloze, Edit, Expand, and End-to-End. Through our analysis of 14 LLMs, we uncover that most LLMs struggle with tasks that demand contextual understanding under limited or under-specified instructions. By augmenting RAVEL with SOTA LLMs as operators, we find that such agentic text synthesis is dominated by the LLM's reasoning capability rather than raw generative capacity. Furthermore, we find that a strong reasoner can guide a weaker generator to yield higher-quality results, whereas the inverse does not hold. Our code and data are available at this link: https://github.com/ZhuoerFeng/RAVEL-Reasoning-Agents-Text-Eval.

Paperid: 375, https://arxiv.org/pdf/2603.00682.pdf GitHub

Abstract:
Collaborative perception empowers autonomous agents to share complementary information and overcome perception limitations. While early fusion offers more perceptual complementarity and is inherently robust to model heterogeneity, its high communication cost has limited its practical deployment, prompting most existing works to favor intermediate or late fusion. To address this, we propose a communication-efficient early Collaborative perception framework that incorporates LiDAR Completion to restore scene completeness under sparse transmission, dubbed as CoLC. Specifically, the CoLC integrates three complementary designs. First, each neighbor agent applies Foreground-Aware Point Sampling (FAPS) to selectively transmit informative points that retain essential structural and contextual cues under bandwidth constraints. The ego agent then employs Completion-Enhanced Early Fusion (CEEF) to reconstruct dense pillars from the received sparse inputs and adaptively fuse them with its own observations, thereby restoring spatial completeness. Finally, the Dense-Guided Dual Alignment (DGDA) strategy enforces semantic and geometric consistency between the enhanced and dense pillars during training, ensuring consistent and robust feature learning. Experiments on both simulated and real-world datasets demonstrate that CoLC achieves superior perception-communication trade-offs and remains robust under heterogeneous model settings. The code is available at https://github.com/CatOneTwo/CoLC.

Paperid: 376, https://arxiv.org/pdf/2603.00656.pdf GitHub

Abstract:
Real-world user requests to LLM agents are often underspecified. Agents must interact to acquire missing information and make correct downstream decisions. However, current multi-turn GRPO-based methods often rely on trajectory-level reward computation, which leads to credit assignment problems and insufficient advantage signals within rollout groups. A feasible approach is to identify valuable interaction turns at a fine granularity to drive more targeted learning. To address this, we introduce InfoPO (Information-Driven Policy Optimization), which frames multi-turn interaction as a process of active uncertainty reduction and computes an information-gain reward that credits turns whose feedback measurably changes the agent's subsequent action distribution compared to a masked-feedback counterfactual. It then combines this signal with task outcomes via an adaptive variance-gated fusion to identify information importance while maintaining task-oriented goal direction. Across diverse tasks, including intent clarification, collaborative coding, and tool-augmented decision making, InfoPO consistently outperforms prompting and multi-turn RL baselines. It also demonstrates robustness under user simulator shifts and generalizes effectively to environment-interactive tasks. Overall, InfoPO provides a principled and scalable mechanism for optimizing complex agent-user collaboration. Code is available at https://github.com/kfq20/InfoPO.

Paperid: 377, https://arxiv.org/pdf/2603.00651.pdf GitHub GitHub

Abstract:
Dataset pruning has been widely studied for 2D images to remove redundancy and accelerate training, while particular pruning methods for 3D data remain largely unexplored. In this work, we study dataset pruning for 3D data, where its observed common long-tail class distribution nature make optimization under conventional evaluation metrics Overall Accuracy (OA) and Mean Accuracy (mAcc) inherently conflicting, and further make pruning particularly challenging. To address this, we formulate pruning as approximating the full-data expected risk with a weighted subset, which reveals two key errors: coverage error from insufficient representativeness and prior-mismatch bias from inconsistency between subset-induced class weights and target metrics. We propose representation-aware subset selection with per-class retention quotas for long-tail coverage, and prior-invariant teacher supervision using calibrated soft labels and embedding-geometry distillation. The retention quota also serves as a switch to control the OA-mAcc trade-off. Extensive experiments on 3D datasets show that our method can improve both metrics across multiple settings while adapting to different downstream preferences. Our code is available at https://github.com/XiaohanZhao123/3D-Dataset-Pruning.

Paperid: 378, https://arxiv.org/pdf/2603.00638.pdf GitHub

Abstract:
Large language models (LLMs) are increasingly adopted as the backbone of recommender systems. However, user-item interactions in real-world scenarios are non-stationary, making preference drift over time inevitable. Existing model update strategies mainly rely on global fine-tuning or pointwise editing, but they face two fundamental challenges: (i) imbalanced update granularity, where global updates perturb behaviors unrelated to the target while pointwise edits fail to capture broader preference shifts; (ii) unstable incremental updates, where repeated edits interfere with prior adaptations, leading to catastrophic forgetting and inconsistent recommendations. To address these issues, we propose Region-Aware Incremental Editing (RAIE), a plug-in framework that freezes the backbone model and performs region-level updates. RAIE first constructs semantically coherent preference regions via spherical k-means in the representation space. It then assigns incoming sequences to regions via confidence-aware gating and performs three localized edit operations - Update, Expand, and Add - to dynamically revise the affected region. Each region is equipped with a dedicated Low-Rank Adaptation (LoRA) module, which is trained only on the region's updated data. During inference, RAIE routes each user sequence to its corresponding region and activates the region-specific adapter for prediction. Experiments on two benchmark datasets under a time-sliced protocol that segments data into Set-up (S), Finetune (F), and Test (T) show that RAIE significantly outperforms state-of-the-art baselines while effectively mitigating forgetting. These results demonstrate that region-aware editing offers an accurate and scalable mechanism for continual adaptation in dynamic recommendation scenarios. Our code is available at https://github.com/fengaogao/RAIE.

Paperid: 379, https://arxiv.org/pdf/2603.00636.pdf GitHub

Abstract:
We propose a retrodictive forecasting paradigm for time series: instead of predicting the future from the past, we identify the future that best explains the observed present via inverse MAP optimization over a Conditional Variational Autoencoder (CVAE). This conditioning is a statistical modeling choice for Bayesian inversion; it does not assert that future events cause past observations. The approach is theoretically grounded in an information-theoretic arrow-of-time measure: the symmetrized Kullback-Leibler divergence between forward and time-reversed trajectory ensembles provides both the conceptual rationale and an operational GO/NO-GO diagnostic for applicability. We implement the paradigm as MAP inference over an inverse CVAE with a learned RealNVP normalizing-flow prior and evaluate it on six time series cases: four synthetic processes with controlled temporal asymmetry and two ERA5 reanalysis datasets (wind speed and solar irradiance). The work makes four contributions: (i) a formal retrodictive inference formulation; (ii) an inverse CVAE architecture; (iii) a model-free irreversibility diagnostic; and (iv) a falsifiable validation protocol with four pre-specified predictions. All pre-specified predictions are empirically supported: the diagnostic correctly classifies all six cases; the learned flow prior improves over an isotropic Gaussian baseline on GO cases; the inverse MAP yields no spurious advantage on time-reversible dynamics; and on irreversible GO cases, it achieves competitive or superior RMSE relative to forward baselines, with a statistically significant 17.7% reduction over a forward MLP on ERA5 solar irradiance. These results provide a structured proof-of-concept that retrodictive forecasting can constitute a viable alternative to conventional forward prediction when statistical time-irreversibility is present and exploitable.

Paperid: 380, https://arxiv.org/pdf/2603.00634.pdf GitHub

Abstract:
Multilingual falsehoods threaten information integrity worldwide, yet detection benchmarks remain confined to English or a few high-resource languages, leaving low-resource linguistic communities without robust defense tools. We introduce BLUFF, a comprehensive benchmark for detecting false and synthetic content, spanning 79 languages with over 202K samples, combining human-written fact-checked content (122K+ samples across 57 languages) and LLM-generated content (79K+ samples across 71 languages). BLUFF uniquely covers both high-resource "big-head" (20) and low-resource "long-tail" (59) languages, addressing critical gaps in multilingual research on detecting false and synthetic content. Our dataset features four content types (human-written, LLM-generated, LLM-translated, and hybrid human-LLM text), bidirectional translation (English$\leftrightarrow$X), 39 textual modification techniques (36 manipulation tactics for fake news, 3 AI-editing strategies for real news), and varying edit intensities generated using 19 diverse LLMs. We present AXL-CoI (Adversarial Cross-Lingual Agentic Chainof-Interactions), a novel multi-agentic framework for controlled fake/real news generation, paired with mPURIFY, a quality filtering pipeline ensuring dataset integrity. Experiments reveal state-of-theart detectors suffer up to 25.3% F1 degradation on low-resource versus high-resource languages. BLUFF provides the research community with a multilingual benchmark, extensive linguistic-oriented benchmark evaluation, comprehensive documentation, and opensource tools to advance equitable falsehood detection. Dataset and code are available at: https://jsl5710.github.io/BLUFF/

Paperid: 381, https://arxiv.org/pdf/2603.00631.pdf GitHub

Abstract:
LiTS is a modular Python framework for LLM reasoning via tree search. It decomposes tree search into three reusable components (Policy, Transition, and RewardModel) that plug into algorithms like MCTS and BFS. A decorator-based registry enables domain experts to extend to new domains by registering components, and algorithmic researchers to implement custom search algorithms. We demonstrate composability on MATH500 (language reasoning), Crosswords (environment planning), and MapEval (tool use), showing that components and algorithms are orthogonal: components are reusable across algorithms within each task type, and algorithms work across all components and domains. We also report a mode-collapse finding: in infinite action spaces, LLM policy diversity (not reward quality) is the bottleneck for effective tree search. A demonstration video is available at https://youtu.be/nRGX43YrR3I. The package is released under the Apache 2.0 license at https://github.com/xinzhel/lits-llm, including installation instructions and runnable examples that enable users to reproduce the demonstrated workflows.

Paperid: 382, https://arxiv.org/pdf/2603.00629.pdf GitHub

Abstract:
Large time series models (LTMs) have emerged as powerful tools for universal forecasting, yet they often struggle with the inherent diversity and nonstationarity of real-world time series data, leading to an unsatisfactory trade-off between forecasting accuracy and generalization. Rather than continually finetuning new LTM instances for each domain, we propose a data-centric framework, time-series adaptive transformation optimization (TATO), that enables a single frozen pre-trained LTM to adapt to diverse downstream domains through an optimally configured transformation pipeline. Specifically, TATO constructs three representative types of transformations, including context slicing, scale normalization, and outlier correction, to help LTMs better align with target domain characteristics. To ensure robustness, we incorporate carefully selected time series augmentations and a two-stage ranking mechanism that filters out pipelines underperforming on specific metrics. Extensive experiments on state-of-the-art LTMs and widely used datasets demonstrate that TATO consistently and significantly improves domain-adaptive forecasting performance, achieving a maximum reduction in MSE of 65.4\% and an average reduction of 13.6\%. Moreover, TATO is highly efficient, typically completing optimization in under 2 minutes, making it practical for real-world deployment. The source code is available at https://github.com/thulab/TATO.

Paperid: 383, https://arxiv.org/pdf/2603.00624.pdf GitHub

Abstract:
Catastrophic forgetting, the tendency of neural networks to forget previously learned knowledge when learning new tasks, has been a major challenge in continual learning (CL). To tackle this challenge, CL methods have been proposed and shown to reduce forgetting. Furthermore, CL models deployed in mission-critical settings can benefit from uncertainty awareness by calibrating their predictions to reliably assess their confidences. However, existing uncertainty-aware continual learning methods suffer from high computational overhead and incompatibility with mainstream replay methods. To address this, we propose idempotent experience replay (IDER), a novel approach based on the idempotent property where repeated function applications yield the same output. Specifically, we first adapt the training loss to make model idempotent on current data streams. In addition, we introduce an idempotence distillation loss. We feed the output of the current model back into the old checkpoint and then minimize the distance between this reprocessed output and the original output of the current model. This yields a simple and effective new baseline for building reliable continual learners, which can be seamlessly integrated with other CL approaches. Extensive experiments on different CL benchmarks demonstrate that IDER consistently improves prediction reliability while simultaneously boosting accuracy and reducing forgetting. Our results suggest the potential of idempotence as a promising principle for deploying efficient and trustworthy continual learning systems in real-world applications.Our code is available at https://github.com/YutingLi0606/Idempotent-Continual-Learning.

Paperid: 384, https://arxiv.org/pdf/2603.00623.pdf GitHub

Abstract:
Agentic systems augment large language models with external tools and iterative decision making, enabling complex tasks such as deep research, function calling, and coding. However, their long and intricate execution traces make failure diagnosis and root cause analysis extremely challenging. Manual inspection does not scale, while directly applying LLMs to raw traces is hindered by input length limits and unreliable reasoning. Focusing solely on final task outcomes further discards critical behavioral information required for accurate issue localization. To address these issues, we propose TraceSIR, a multi-agent framework for structured analysis and reporting of agentic execution traces. TraceSIR coordinates three specialized agents: (1) StructureAgent, which introduces a novel abstraction format, TraceFormat, to compress execution traces while preserving essential behavioral information; (2) InsightAgent, which performs fine-grained diagnosis including issue localization, root cause analysis, and optimization suggestions; (3) ReportAgent, which aggregates insights across task instances and generates comprehensive analysis reports. To evaluate TraceSIR, we construct TraceBench, covering three real-world agentic scenarios, and introduce ReportEval, an evaluation protocol for assessing the quality and usability of analysis reports aligned with industry needs. Experiments show that TraceSIR consistently produces coherent, informative, and actionable reports, significantly outperforming existing approaches across all evaluation dimensions. Our project and video are publicly available at https://github.com/SHU-XUN/TraceSIR.

Paperid: 385, https://arxiv.org/pdf/2603.00621.pdf GitHub

Abstract:
Research in CDCR remains fragmented due to heterogeneous dataset formats, varying annotation standards, and the predominance of the CDCR definition as the event coreference resolution (ECR). To address these challenges, we introduce uCDCR, a unified dataset that consolidates diverse publicly available English CDCR corpora across various domains into a consistent format, which we analyze with standardized metrics and evaluation protocols. uCDCR incorporates both entity and event coreference, corrects known inconsistencies, and enriches datasets with missing attributes to facilitate reproducible research. We establish a cohesive framework for fair, interpretable, and cross-dataset analysis in CDCR and compare the datasets on their lexical properties, e.g., lexical composition of the annotated mentions, lexical diversity and ambiguity metrics, discuss the annotation rules and principles that lead to high lexical diversity, and examine how these metrics influence performance on the same-head-lemma baseline. Our dataset analysis shows that ECB+, the state-of-the-art benchmark for CDCR, has one of the lowest lexical diversities, and its CDCR complexity, measured by the same-head-lemma baseline, lies in the middle among all uCDCR datasets. Moreover, comparing document and mention distributions between ECB+ and uCDCR shows that using all uCDCR datasets for model training and evaluation will improve the generalizability of CDCR models. Finally, the almost identical performance on the same-head-lemma baseline, separately applied to events and entities, shows that resolving both types is a complex task and should not be steered toward ECR alone. The uCDCR dataset is available at https://huggingface.co/datasets/AnZhu/uCDCR, and the code for parsing, analyzing, and scoring the dataset is available at https://github.com/anastasia-zhukova/uCDCR.

Paperid: 386, https://arxiv.org/pdf/2603.00618.pdf GitHub

Abstract:
Multi-domain graph pre-training integrates knowledge from diverse domains to enhance performance in the target domains, which is crucial for building graph foundation models. Despite initial success, existing solutions often fall short of answering a fundamental question: how is knowledge integrated or transferred across domains? This theoretical limitation motivates us to rethink the consistency and transferability between model pre-training and domain adaptation. In this paper, we propose a fresh Riemannian geometry perspective, whose core idea is to merge any graph dataset into a unified, smooth Riemannian manifold, enabling a systematic understanding of knowledge integration and transfer. To achieve this, our key contribution is the theoretical establishment of neural manifold gluing, which first characterizes local geometry using an adaptive orthogonal frame and then "glues" the local pieces together into a coherent whole. Building on this theory, we present the GraphGlue framework, which supports batched pre-training with EMA prototyping and provides a transferability measure based on geometric consistence. Extensive experiments demonstrate its superior performance across diverse graph domains. Moreover, we empirically validated GraphGlue's geometric scaling law, showing that larger quantities of datasets improve model transferability by producing a smoother manifold. Codes are available at https://github.com/RiemannGraph/GraphGlue.

Paperid: 387, https://arxiv.org/pdf/2603.00615.pdf GitHub

Abstract:
The performance of robotic imitation learning is fundamentally limited by data quality and training strategies. Prevalent sampling strategies on RLBench suffer from severe keyframe redundancy and imbalanced temporal distribution, leading to inefficient memory usage and unstable optimization. Moreover, reprojecting point clouds onto multi-view images with a black background--while more efficient than voxel-based methods--often causes dark objects to be indistinguishable and hard to manipulate. In this work, we propose a novel holistic framework that significantly improves both model performance and training efficiency. First, we redesign and optimize the keyframe sampling strategy, reducing memory consumption by 80% and accelerating training speed by 5x. Second, we augment the model with a color inversion projection branch--a simple yet effective module that resolves the ambiguity of dark objects. Finally, we propose a task-guided mixup technique that dynamically fuses point clouds and action heatmaps according to task instructions, greatly improving robustness to distractors and performance in multi-goal scenarios. Extensive experiments demonstrate that our method achieves state-of-the-art performance with a 90.5% success rate on RLBench and 68.8% on the COLOSSEUM benchmark under challenging interference conditions. Our code and checkpoints are available at https://github.com/PuFanqi23/TGM-VLA.

Paperid: 388, https://arxiv.org/pdf/2603.00611.pdf GitHub

Abstract:
Recently, Spectral Compressive Imaging (SCI) has achieved remarkable success, unlocking significant potential for dynamic spectral vision. However, existing reconstruction methods, primarily image-based, suffer from two limitations: (i) Encoding process masks spatial-spectral features, leading to uncertainty in reconstructing missing information from single compressed measurements, and (ii) The frame-by-frame reconstruction paradigm fails to ensure temporal consistency, which is crucial in the video perception. To address these challenges, this paper seeks to advance spectral reconstruction from the image level to the video level, leveraging the complementary features and temporal continuity across adjacent frames in dynamic scenes. Initially, we construct the first high-quality dynamic hyperspectral image dataset (DynaSpec), comprising 30 sequences obtained through frame-scanning acquisition. Subsequently, we propose the Propagation-Guided Spectral Video Reconstruction Transformer (PG-SVRT), which employs a spatial-then-temporal attention to effectively reconstruct spectral features from abundant video information, while using a bridged token to reduce computational complexity. Finally, we conduct simulation experiments to assess the performance of four SCI systems, and construct a DD-CASSI prototype for real-world data collection and benchmarking. Extensive experiments demonstrate that PG-SVRT achieves superior performance in reconstruction quality, spectral fidelity, and temporal consistency, while maintaining minimal FLOPs. Project page: https://github.com/nju-cite/DynaSpec

Paperid: 389, https://arxiv.org/pdf/2603.00609.pdf GitHub

Abstract:
Collaborative perception leverages data exchange among multiple agents to enhance overall perception capabilities. However, heterogeneity across agents introduces domain gaps that hinder collaboration, and this is further exacerbated by an underexplored issue: modality isolation. It arises when multiple agents with different modalities never co-occur in any training data frame, enlarging cross-modal domain gaps. Existing alignment methods rely on supervision from spatially overlapping observations, thus fail to handle modality isolation. To address this challenge, we propose CodeAlign, the first efficient, co-occurrence-free alignment framework that smoothly aligns modalities via cross-modal feature-code-feature(FCF) translation. The key idea is to explicitly identify the representation consistency through codebook, and directly learn mappings between modality-specific feature spaces, thereby eliminating the need for spatial correspondence. Codebooks regularize feature spaces into code spaces, providing compact yet expressive representations. With a prepared code space for each modality, CodeAlign learns FCF translations that map features to the corresponding codes of other modalities, which are then decoded back into features in the target code space, enabling effective alignment. Experiments show that, when integrating three modalities, CodeAlign requires only 8% of the training parameters of prior alignment methods, reduces communication load by 1024x, and achieves state-of-the-art perception performance on both OPV2V and DAIR-V2X dataset. Code will be released on https://github.com/cxliu0314/CodeAlign.

Paperid: 390, https://arxiv.org/pdf/2603.00606.pdf GitHub

Abstract:
Accurate 3D hand pose and pressure sensing is essential for immersive human-computer interaction, yet simultaneously achieving both in mobile scenarios remains a significant challenge. We present WristPP, a camera-based wrist-worn system that estimates 3D hand pose and per-vertex pressure from a single wide-FOV RGB frame in real time. A Vision Transformer (ViT) backbone with joint-aligned tokens predicts Hand-VQVAE codebook indices for mesh recovery, while an extrinsics-conditioned branch jointly estimates per-vertex pressure. On a self-collected dataset of 133,000 frames (20 subjects; 48 on-plane and 28 mid-air gestures), WristPP attains a Mean Per-Joint Position Error (MPJPE) of 2.9 mm, Contact IoU of 0.712, Volumetric IoU of 0.618, and foreground pressure MAE of 10.4 g. Across three user studies, WristPP delivers touchpad-level efficiency in mid-air pointing and robust multi-finger pressure control on an uninstrumented desktop. In a real-world large-display Whac-A-Mole task, WristPP also enables higher success ratio and lower arm fatigue than head-mounted camera-based baselines. These results position WristPP as an effective, mobile solution for versatile pose- and pressure-based interaction. Website: https://zhenqis123.github.io/WristPP/.

Paperid: 391, https://arxiv.org/pdf/2603.00604.pdf GitHub

Abstract:
High-quality pixel-level annotations are essential for the semantic segmentation of remote sensing imagery. However, such labels are expensive to obtain and often affected by noise due to the labor-intensive and time-consuming nature of pixel-wise annotation, which makes it challenging for human annotators to label every pixel accurately. Annotation errors can significantly degrade the performance and robustness of modern segmentation models, motivating the need for reliable mechanisms to identify and quantify noisy training samples. This paper introduces a novel Data-Centric benchmark, together with a novel, publicly available dataset and two techniques for identifying, quantifying, and ranking training samples according to their level of label noise in remote sensing semantic segmentation. Such proposed methods leverage complementary strategies based on model uncertainty, prediction consistency, and representation analysis, and consistently outperform established baselines across a range of experimental settings. The outcomes of this work are publicly available at https://github.com/keillernogueira/label_noise_segmentation.

Paperid: 392, https://arxiv.org/pdf/2603.00601.pdf GitHub

Abstract:
AI code agents excel at isolated tasks yet struggle with multi-file software engineering requiring architectural understanding. We introduce Theory of Code Space (ToCS), a benchmark that evaluates whether agents can construct, maintain, and update coherent architectural beliefs during codebase exploration. Agents explore procedurally generated codebases under partial observability -- opening files under a budget -- and periodically externalize their belief state as structured JSON, producing a time-series of architectural understanding. Three findings emerge from experiments with four baselines and six frontier LLMs. First, the Active-Passive Gap is model-dependent: one model builds better maps through active exploration than from seeing all files at once, while another shows the opposite -- revealing that active exploration is itself a non-trivial capability absent from some models. Second, retaining structured belief maps in context acts as self-scaffolding for some models but not others, showing that the mechanism is model-dependent. Third, belief state maintenance varies dramatically: a smaller model maintains perfectly stable beliefs across probes while its larger sibling suffers catastrophic belief collapse -- forgetting previously-discovered components between probes. We release ToCS as open-source software. Code: https://github.com/che-shr-cat/tocs

Paperid: 393, https://arxiv.org/pdf/2603.00597.pdf GitHub

Abstract:
Inertial Odometry (IO) has gained attention in quadrotor applications due to its sole reliance on inertial measurement units (IMUs), attributed to its lightweight design, low cost, and robust performance across diverse environments. However, most existing learning-based inertial odometry systems for quadrotors either use only IMU data or include additional dynamics-related inputs such as thrust, but still lack a principled formulation of the underlying physical model to be learned. This lack of interpretability hampers the model's ability to generalize and often limits its accuracy. In this work, we approach the inertial odometry learning problem from a different perspective. Inspired by the aerodynamics model and IMU measurement model, we identify the key physical quantity--rotor speed measurements required for inertial odometry and design a transformer-based inertial odometry. By incorporating rotor speed measurements, the proposed model improves velocity prediction accuracy by 36.9%. Furthermore, the transformer architecture more effectively exploits temporal dependencies for denoising and aerodynamics modeling, yielding an additional 22.4% accuracy gain over previous results. To support evaluation, we also provide a real-world quadrotor flight dataset capturing IMU measurements and rotor speed for high-speed motion. Finally, combined with an uncertainty-aware extended Kalman filter (EKF), our framework is validated across multiple datasets and real-time systems, demonstrating superior accuracy, generalization, and real-time performance. We share the code and data to promote further research (https://github.com/SJTU-ViSYS-team/AI-IO).

Paperid: 394, https://arxiv.org/pdf/2603.00592.pdf GitHub

Abstract:
Vision-Language-Action (VLA) models achieve over 95% success on standard benchmarks. However, through systematic experiments, we find that current state-of-the-art VLA models largely ignore language instructions. Prior work lacks: (1) systematic semantic perturbation diagnostics, (2) a benchmark that forces language understanding by design, and (3) linguistically diverse training data. This paper constructs the LangGap benchmark, based on a four-dimensional semantic perturbation method -- varying instruction semantics while keeping the tabletop layout fixed -- revealing language understanding deficits in π0.5. Existing benchmarks like LIBERO assign only one task per layout, underutilizing available objects and target locations; LangGap fully diversifies pick-and-place tasks under identical layouts, forcing models to truly understand language. Experiments show that targeted data augmentation can partially close the language gap -- success rate improves from 0% to 90% with single-task training, and 0% to 28% with multi-task training. However, as semantic diversity of extended tasks increases, model learning capacity proves severely insufficient; even trained tasks perform poorly. This reveals a fundamental challenge for VLA models in understanding diverse language instructions -- precisely the long-term value of LangGap.

Paperid: 395, https://arxiv.org/pdf/2603.00585.pdf GitHub

Abstract:
Recent advances in video generation have opened new avenues for macroscopic simulation of complex dynamic systems, but their application to microscopic phenomena remains largely unexplored. Microscale simulation holds great promise for biomedical applications such as drug discovery, organ-on-chip systems, and disease mechanism studies, while also showing potential in education and interactive visualization. In this work, we introduce MicroWorldBench, a multi-level rubric-based benchmark for microscale simulation tasks. MicroWorldBench enables systematic, rubric-based evaluation through 459 unique expert-annotated criteria spanning multiple microscale simulation task (e.g., organ-level processes, cellular dynamics, and subcellular molecular interactions) and evaluation dimensions (e.g., scientific fidelity, visual quality, instruction following). MicroWorldBench reveals that current SOTA video generation models fail in microscale simulation, showing violations of physical laws, temporal inconsistency, and misalignment with expert criteria. To address these limitations, we construct MicroSim-10K, a high-quality, expert-verified simulation dataset. Leveraging this dataset, we train MicroVerse, a video generation model tailored for microscale simulation. MicroVerse can accurately reproduce complex microscale mechanism. Our work first introduce the concept of Micro-World Simulation and present a proof of concept, paving the way for applications in biology, education, and scientific visualization. Our work demonstrates the potential of educational microscale simulations of biological mechanisms. Our data and code are publicly available at https://github.com/FreedomIntelligence/MicroVerse

Paperid: 396, https://arxiv.org/pdf/2603.00582.pdf GitHub

Abstract:
While Large Language Models (LLMs) have demonstrated proficiency in Deep Research or Wide Search, their capacity to solve highly complex questions-those requiring long-horizon planning, massive evidence gathering, and synthesis across heterogeneous sources-remains largely unexplored. We introduce Super Research, a task for complex autonomous research tasks that integrates (i) structured decomposition into a research plan, (ii) super wide retrieval for diverse perspectives, and (iii) super deep investigation to resolve uncertainties through iterative queries. To evaluate this capability, we curated a benchmark of 300 expert-written questions across diverse domains, each requiring up to 100+ retrieval steps and 1,000+ web pages to reconcile conflicting evidence. Super Research produces verifiable reports with fine-grained citations and intermediate artifacts (e.g., outlines and tables) to ensure traceable reasoning. Furthermore, we present a graph-anchored auditing protocol that evaluates Super Research along five dimensions: Coverage, Logical Consistency, Report Utility, Objectivity and Citation Health. While super-complex questions may be infrequent in standard applications, Super Research serves as a critical ceiling evaluation and stress test for LLM capabilities. A model's proficiency within Super Research acts as a powerful proxy for its general research competence; success here suggests the robustness necessary to navigate nearly any subordinate research task. Leaderboard is available at: https://cnsdqd-dyb.github.io/Super-Research-Benchmark/

Paperid: 397, https://arxiv.org/pdf/2603.00576.pdf GitHub

Abstract:
Symbolic music generation is a challenging task in multimedia generation, involving long sequences with hierarchical temporal structures, long-range dependencies, and fine-grained local details. Though recent diffusion-based models produce high quality generations, they tend to suffer from high training and inference costs with long symbolic sequences due to iterative denoising and sequence-length-related costs. To deal with such problem, we put forth a diffusing strategy named SMDIM to combine efficient global structure construction and light local refinement. SMDIM uses structured state space models to capture long range musical context at near linear cost, and selectively refines local musical details via a hybrid refinement scheme. Experiments performed on a wide range of symbolic music datasets which encompass various Western classical music, popular music and traditional folk music show that the SMDIM model outperforms the other state-of-the-art approaches on both the generation quality and the computational efficiency, and it has robust generalization to underexplored musical styles. These results show that SMDIM offers a principled solution for long-sequence symbolic music generation, including associated attributes that accompany the sequences. We provide a project webpage with audio examples and supplementary materials at https://3328702107.github.io/smdim-music/.

Paperid: 398, https://arxiv.org/pdf/2603.00565.pdf GitHub

Abstract:
Multimodal Large Language Models (MLLMs) have achieved remarkable performance but remain vulnerable to jailbreak attacks that can induce harmful content and undermine their secure deployment. Previous studies have shown that introducing additional inference steps, which disrupt security attention, can make MLLMs more susceptible to being misled into generating malicious content. However, these methods rely on single-image masking or isolated visual cues, which only modestly extend reasoning paths and thus achieve limited effectiveness, particularly against strongly aligned commercial closed-source models. To address this problem, in this paper, we propose Multi-Image Dispersion and Semantic Reconstruction (MIDAS), a multimodal jailbreak framework that decomposes harmful semantics into risk-bearing subunits, disperses them across multiple visual clues, and leverages cross-image reasoning to gradually reconstruct the malicious intent, thereby bypassing existing safety mechanisms. The proposed MIDAS enforces longer and more structured multi-image chained reasoning, substantially increases the model's reliance on visual cues while delaying the exposure of malicious semantics and significantly reducing the model's security attention, thereby improving the performance of jailbreak against advanced MLLMs. Extensive experiments across different datasets and MLLMs demonstrate that the proposed MIDAS outperforms state-of-the-art jailbreak attacks for MLLMs and achieves an average attack success rate of 81.46% across 4 closed-source MLLMs. Our code is available at this [link](https://github.com/Winnie-Lian/MIDAS).

Paperid: 399, https://arxiv.org/pdf/2603.00543.pdf GitHub

Abstract:
Pansharpening aims to generate high-resolution multi-spectral images by fusing the spatial detail of panchromatic images with the spectral richness of low-resolution MS data. However, most existing methods are evaluated under limited, low-resolution settings, limiting their generalization to real-world, high-resolution scenarios. To bridge this gap, we systematically investigate the data, algorithmic, and computational challenges of cross-scale pansharpening. We first introduce PanScale, the first large-scale, cross-scale pansharpening dataset, accompanied by PanScale-Bench, a comprehensive benchmark for evaluating generalization across varying resolutions and scales. To realize scale generalization, we propose ScaleFormer, a novel architecture designed for multi-scale pansharpening. ScaleFormer reframes generalization across image resolutions as generalization across sequence lengths: it tokenizes images into patch sequences of the same resolution but variable length proportional to image scale. A Scale-Aware Patchify module enables training for such variations from fixed-size crops. ScaleFormer then decouples intra-patch spatial feature learning from inter-patch sequential dependency modeling, incorporating Rotary Positional Encoding to enhance extrapolation to unseen scales. Extensive experiments show that our approach outperforms SOTA methods in fusion quality and cross-scale generalization. The datasets and source code are available at https://github.com/caoke-963/ScaleFormer.

Paperid: 400, https://arxiv.org/pdf/2603.00535.pdf GitHub

Abstract:
Cone-beam CT (CBCT) is routinely acquired in radiotherapy but suffers from severe artifacts and unreliable Hounsfield Unit (HU) values, limiting its direct use for dose calculation. Synthetic CT (sCT) generation from CBCT is therefore an important task, yet paired CBCT--CT data are often unavailable or unreliable due to temporal gaps, anatomical variation, and registration errors. In this work, we introduce rectified flow (RF) into unpaired CBCT-to-CT translation in medical imaging. Although RF is theoretically compatible with unpaired learning through distribution-level coupling and deterministic transport, its practical effectiveness under small medical datasets and limited batch sizes remains underexplored. Direct application with random or batch-local pseudo pairing can produce unstable supervision due to semantically mismatched endpoint samples. To address this challenge, we propose Retrieval-Augmented Flow Matching (RAFM), which adapts RF to the medical setting by constructing retrieval-guided pseudo pairs using a frozen DINOv3 encoder and a global CT memory bank. This strategy improves empirical coupling quality and stabilizes unpaired flow-based training. Experiments on SynthRAD2023 under a strict subject-level true-unpaired protocol show that RAFM outperforms existing methods across FID, MAE, SSIM, PSNR, and SegScore. The code is available at https://github.com/HiLab-git/RAFM.git.

Paperid: 401, https://arxiv.org/pdf/2603.00519.pdf GitHub

Abstract:
Diffusion models have achieved remarkable success in generative AI, yet their computational efficiency remains a significant challenge, particularly for Diffusion Transformers (DiTs) requiring intensive full-attention computation. While existing acceleration approaches focus on content-agnostic uniform optimization strategies, we observe that different regions in generated content exhibit heterogeneous convergence patterns during the denoising process. We present Jano, a training-free framework that leverages this insight for efficient region-aware generation. Jano introduces an early-stage complexity recognition algorithm that accurately identifies regional convergence requirements within initial denoising steps, coupled with an adaptive token scheduling runtime that optimizes computational resource allocation. Through comprehensive evaluation on state-of-the-art models, Jano achieves substantial acceleration (average 2.0 times speedup, up to 2.4 times) while preserving generation quality. Our work challenges conventional uniform processing assumptions and provides a practical solution for accelerating large-scale content generation. The source code of our implementation is available at https://github.com/chen-yy20/Jano.

Paperid: 402, https://arxiv.org/pdf/2603.00516.pdf GitHub

Abstract:
Federated Instruction Tuning (FIT) enables collaborative instruction tuning of large language models across multiple organizations (clients) in a cross-silo setting without requiring the sharing of private instructions. Recent findings on natural backdoors and the existing training data collection method suggest that poisoned samples may be pervasive and inadvertently embedded in real-world datasets, potentially distributed across all clients, even if the clients are benign. This work systematically examine this threat in FIT, demonstrating that existing defenses are ineffective when poisoned data is interspersed among all clients. Addressing this challenge entails two major difficulties: identifying the distinctive characteristics of poisoned samples at each client and enabling collaborative defense when some clients are heavily dominated by poisoned samples. To address these difficulties, we identify gradients in the frequency domain as a robust signal to distinguish poisoned data. We further propose a global secondary clustering mechanism that facilitates collaborative identification of poisoned samples across clients. In summary, this paper introduces ProtegoFed, the first backdoor-free FIT framework that accurately detects, removes, and even purifies interspersed poisoned data across clients during the training. Experimental results on four FL datasets show that ProtegoFed identifies $92.00\% \sim 100.00\%$ of poisoned samples, reduces the attack success rate to almost zero, and maintains utility on the main task. Code is available at https://github.com/dongdongzhaoUP/ProtegoFed.

Paperid: 403, https://arxiv.org/pdf/2603.00515.pdf GitHub

Abstract:
Humans are born with vision-based 4D spatial-temporal intelligence, which enables us to perceive and reason about the evolution of 3D space over time from purely visual inputs. Despite its importance, this capability remains a significant bottleneck for current multimodal large language models (MLLMs). To tackle this challenge, we introduce MLLM-4D, a comprehensive framework designed to bridge the gaps in training data curation and model post-training for spatiotemporal understanding and reasoning. On the data front, we develop a cost-efficient data curation pipeline that repurposes existing stereo video datasets into high-quality 4D spatiotemporal instructional data. This results in the MLLM4D-2M and MLLM4D-R1-30k datasets for Supervised Fine-Tuning (SFT) and Reinforcement Fine-Tuning (RFT), alongside MLLM4D-Bench for comprehensive evaluation. Regarding model training, our post-training strategy establishes a foundational 4D understanding via SFT and further catalyzes 4D reasoning capabilities by employing Group Relative Policy Optimization (GRPO) with specialized Spatiotemporal Chain of Thought (ST-CoT) prompting and Spatiotemporal reward functions (ST-reward) without involving the modification of architecture. Extensive experiments demonstrate that MLLM-4D achieves state-of-the-art spatial-temporal understanding and reasoning capabilities from purely 2D RGB inputs. Project page: https://github.com/GVCLab/MLLM-4D.

Paperid: 404, https://arxiv.org/pdf/2603.00510.pdf GitHub

Abstract:
Multimodal large language models (MLLMs) project visual tokens into the embedding space of language models, yet the internal structuring and processing of visual semantics remain poorly understood. In this work, we introduce a two-fold analytical framework featuring a novel probing tool, $\textbf{EmbedLens}$, to conduct a fine-grained analysis. We uncover a pronounced semantic sparsity at the input level: visual tokens consistently partition into sink, dead, and alive categories. Remarkably, only the alive tokens, comprising $\approx60\%$ of the total input, carry image-specific meaning. Furthermore, using a targeted patch-compression benchmark, we demonstrate that these alive tokens already encode rich, fine-grained cues (e.g., objects, colors, and OCR) prior to entering the LLM. Internal visual computations (such as visual attention and feed-forward networks) are redundant for most standard tasks. For the small subset of highly vision-centric tasks that actually benefit from internal processing, we reveal that alive tokens naturally align with intermediate LLM layers rather than the initial embedding space, indicating that shallow-layer processing is unnecessary and that direct mid-layer injection is both sufficient. Ultimately, our findings provide a unified mechanistic view of visual token processing, paving the way for more efficient and interpretable MLLM architectures through selective token pruning, minimized visual computation, and mid-layer injection. The code is released at: https://github.com/EIT-NLP/EmbedLens.

Paperid: 405, https://arxiv.org/pdf/2603.00501.pdf GitHub

Abstract:
The integration of large language models (LLMs) into wireless networks has sparked growing interest in building autonomous AI agents for wireless tasks. However, existing approaches rely heavily on manually crafted prompts and static agentic workflows, a process that is labor-intensive, unscalable, and often suboptimal. In this paper, we propose WirelessAgent++, a framework that automates the design of agentic workflows for various wireless tasks. By treating each workflow as an executable code composed of modular operators, WirelessAgent++ casts agent design as a program search problem and solves it with a domain-adapted Monte Carlo Tree Search (MCTS) algorithm. Moreover, we establish WirelessBench, a standardized multi-dimensional benchmark suite comprising Wireless Communication Homework (WCHW), Network Slicing (WCNS), and Mobile Service Assurance (WCMSA), covering knowledge reasoning, code-augmented tool use, and multi-step decision-making. Experiments demonstrate that \wap{} autonomously discovers superior workflows, achieving test scores of $78.37\%$ (WCHW), $90.95\%$ (WCNS), and $97.07\%$ (WCMSA), with a total search cost below $\$ 5$ per task. Notably, our approach outperforms state-of-the-art prompting baselines by up to $31\%$ and general-purpose workflow optimizers by $11.1\%$, validating its effectiveness in generating robust, self-evolving wireless agents. The code is available at https://github.com/jwentong/WirelessAgent-R2.

Paperid: 406, https://arxiv.org/pdf/2603.00487.pdf GitHub

Abstract:
Gentzen's 1936 proof of the consistency of Peano Arithmetic was a significant result in the foundations of mathematics. We provide here a modified version of the proof, based on Gödel's reformulation, and including additional details and minor corrections which are necessary to definitively prove the well-foundedness of the cut-elimination argument in a constructive environment. All results have been verified using the Coq theorem prover. NOTE TO READERS 26 February 2026: this is a draft which we had intended to submit to the Journal of Automated Reasoning with no particular time-line in our minds as the work was completed as part of Aaron's honours project at ANU in 2023. For that reason, we have used the Springer style files. We are putting it on arxiv as there appears to be some interest in this work as indicated by a post to https://proofassistants.stackexchange.com/questions/6462/how-far-is-gentzens-consistency-proof-of-peano-arithmetic-from-being-formalized in early February 2026. The Coq code is available here: https://github.com/aarondroidbryce/Gentzen/tree/master

Paperid: 407, https://arxiv.org/pdf/2603.00486.pdf GitHub

Abstract:
Since Transformers are introduced into vision architectures, their quadratic complexity has always been a significant issue that many research efforts aim to address. A representative approach involves grouping tokens, performing self-attention calculations within each group, or pooling the tokens within each group into a single token. To this end, various carefully designed grouping strategies have been proposed to enhance the performance of Vision Transformers. Here, we pose the following questions: \textbf{Are these carefully designed grouping methods truly necessary? Is there a simpler and more unified token grouping method that can replace these diverse methods?} Therefore, we propose the random grouping strategy, which involves a simple and fast random grouping strategy for vision tokens. We validate this approach on multiple baselines, and experiments show that random grouping almost outperforms all other grouping methods. When transferred to downstream tasks, such as object detection, random grouping demonstrates even more pronounced advantages. In response to this phenomenon, we conduct a detailed analysis of the advantages of random grouping from multiple perspectives and identify several crucial elements for the design of grouping strategies: positional information, head feature diversity, global receptive field, and fixed grouping pattern. We demonstrate that as long as these four conditions are met, vision tokens require only an extremely simple grouping strategy to efficiently and effectively handle various visual tasks. We also validate the effectiveness of our proposed random method across multiple modalities, including visual tasks, point cloud processing, and vision-language models. Code will be available at https://github.com/qhfan/random.

Paperid: 408, https://arxiv.org/pdf/2603.00483.pdf GitHub

Abstract:
Recent text-to-image (T2I) diffusion models achieve remarkable realism, yet faithful prompt-image alignment remains challenging, particularly for complex prompts with multiple objects, relations, and fine-grained attributes. Existing training-free inference-time scaling methods rely on fixed iteration budgets that cannot adapt to prompt difficulty, while reflection-tuned models require carefully curated reflection datasets and extensive joint fine-tuning of diffusion and vision-language models, often overfitting to reflection paths data and lacking transferability across models. We introduce RAISE (Requirement-Adaptive Self-Improving Evolution), a training-free, requirement-driven evolutionary framework for adaptive T2I generation. RAISE formulates image generation as a requirement-driven adaptive scaling process, evolving a population of candidates at inference time through a diverse set of refinement actions-including prompt rewriting, noise resampling, and instructional editing. Each generation is verified against a structured checklist of requirements, enabling the system to dynamically identify unsatisfied items and allocate further computation only where needed. This achieves adaptive test-time scaling that aligns computational effort with semantic query complexity. On GenEval and DrawBench, RAISE attains state-of-the-art alignment (0.94 overall GenEval) while incurring fewer generated samples (reduced by 30-40%) and VLM calls (reduced by 80%) than prior scaling and reflection-tuned baselines, demonstrating efficient, generalizable, and model-agnostic multi-round self-improvement. Code is available at https://github.com/LiyaoJiang1998/RAISE.

Paperid: 409, https://arxiv.org/pdf/2603.00479.pdf GitHub

Abstract:
Automated radiology report generation is key for reducing radiologist workload and improving diagnostic consistency, yet generating accurate reports for 3D medical imaging remains challenging. Existing vision-language models face two limitations: they do not leverage segmentation-pretrained encoders, and they inject visual features only at the input layer of language models, losing multi-scale information. We propose U-VLM, which enables hierarchical vision-language modeling in both training and architecture: (1) progressive training from segmentation to classification to report generation, and (2) multi-layer visual injection that routes U-Net encoder features to corresponding language model layers. Each training stage can leverage different datasets without unified annotations. U-VLM achieves state-of-the-art performance on CT-RATE (F1: 0.414 vs 0.258, BLEU-mean: 0.349 vs 0.305) and AbdomenAtlas 3.0 (F1: 0.624 vs 0.518 for segmentation-based detection) using only a 0.1B decoder trained from scratch, demonstrating that well-designed vision encoder pretraining outweighs the benefits of 7B+ pre-trained language models. Ablation studies show that progressive pretraining significantly improves F1, while multi-layer injection improves BLEU-mean. Code is available at https://github.com/yinghemedical/U-VLM.

Paperid: 410, https://arxiv.org/pdf/2603.00478.pdf GitHub

Abstract:
Few-shot transfer has been revolutionized by stronger pre-trained models and improved adaptation algorithms.However, there lacks a unified, rigorous evaluation protocol that is both challenging and realistic for real-world usage. In this work, we establish FEWTRANS, a comprehensive benchmark containing 10 diverse datasets, and propose the Hyperparameter Ensemble (HPE) protocol to overcome the "validation set illusion" in data-scarce regimes. Our empirical findings demonstrate that the choice of pre-trained model is the dominant factor for performance, while many sophisticated transfer methods offer negligible practical advantages over a simple full-parameter fine-tuning baseline. To explain this surprising effectiveness, we provide an in-depth mechanistic analysis showing that full fine-tuning succeeds via distributed micro-adjustments and more flexible reshaping of high-level semantic presentations without suffering from overfitting. Additionally, we quantify the performance collapse of multimodal models in specialized domains as a result of linguistic rarity using adjusted Zipf frequency scores. By releasing FEWTRANS, we aim to provide a rigorous "ruler" to streamline reproducible advances in few-shot transfer learning research. We make the FEWTRANS benchmark publicly available at https://github.com/Frankluox/FewTrans.

Paperid: 411, https://arxiv.org/pdf/2603.00466.pdf GitHub GitHub

Abstract:
Despite impressive progress in video generation, existing models remain limited to surface-level plausibility, lacking a coherent and unified understanding of the world. Prior approaches typically incorporate only a single form of world-related knowledge or rely on rigid alignment strategies to introduce additional knowledge. However, aligning the single world knowledge is insufficient to constitute a world model that requires jointly modeling multiple heterogeneous dimensions (e.g., physical commonsense, 3D and temporal consistency). To address this limitation, we introduce \textbf{DreamWorld}, a unified framework that integrates complementary world knowledge into video generators via a \textbf{Joint World Modeling Paradigm}, jointly predicting video pixels and features from foundation models to capture temporal dynamics, spatial geometry, and semantic consistency. However, naively optimizing these heterogeneous objectives can lead to visual instability and temporal flickering. To mitigate this issue, we propose \textit{Consistent Constraint Annealing (CCA)} to progressively regulate world-level constraints during training, and \textit{Multi-Source Inner-Guidance} to enforce learned world priors at inference. Extensive evaluations show that DreamWorld improves world consistency, outperforming Wan2.1 by 2.26 points on VBench. Code will be made publicly available at \href{https://github.com/ABU121111/DreamWorld}{\textcolor{mypink}{\textbf{Github}}}.

Paperid: 412, https://arxiv.org/pdf/2603.00455.pdf GitHub

Abstract:
In this work, we present a test-driven, agentic framework for synthesizing a deployable low-level robot controller for navigation tasks. Given a 2D map with an image of an ultrasonic sensor-based robot, or a 3D robotic simulation environment, our framework iteratively refines the generated controller code using diagnostic feedback from structured test suites to achieve task success. We propose a dual-tier repair strategy to refine the generated code that alternates between prompt-level refinement and direct code editing. We evaluate the approach across 2D navigation tasks and 3D navigation in the Webots simulator. Experimental results show that test-driven synthesis substantially improves controller reliability and robustness over one-shot controller generation, especially when the initial prompt is underspecified. The source code and demonstration videos are available at: https://shivanshutripath.github.io/robotic_controller.github.io.

Paperid: 413, https://arxiv.org/pdf/2603.00439.pdf GitHub

Abstract:
Computer-Aided Design (CAD) generative modeling has a strong and long-term application in the industry. Recently, the parametric CAD sequence as the design logic of an object has been widely mined by sequence models. However, the industrial CAD models, especially in component objects, are fine-grained and complex, requiring a longer parametric CAD sequence to define. To address the problem, we introduce Mamba-CAD, a self-supervised generative modeling for complex CAD models in the industry, which can model on a longer parametric CAD sequence. Specifically, we first design an encoder-decoder framework based on a Mamba architecture and pair it with a CAD reconstruction task for pre-training to model the latent representation of CAD models; and then we utilize the learned representation to guide a generative adversarial network to produce the fake representation of CAD models, which would be finally recovered into parametric CAD sequences via the decoder of MambaCAD. To train Mamba-CAD, we further create a new dataset consisting of 77,078 CAD models with longer parametric CAD sequences. Comprehensive experiments are conducted to demonstrate the effectiveness of our model under various evaluation metrics, especially in the generation length of valid parametric CAD sequences. The code and dataset can be achieved from https://github.com/Sunny-Hack/Code-for-Mamba-CAD-AAAI-2025-.

Paperid: 414, https://arxiv.org/pdf/2603.00433.pdf GitHub

Abstract:
Executing multiple tasks simultaneously in medical image analysis, including segmentation, classification, detection, and regression, often introduces significant challenges regarding model generalizability and the optimization of shared feature representations. While Vision Foundation Models (VFMs) provide powerful general representations, full fine-tuning on limited medical data is prone to overfitting and incurs high computational costs. Moreover, existing parameter-efficient fine-tuning approaches typically adopt task-agnostic adaptation protocols, overlooking both task-specific mechanisms and the varying sensitivity of model layers during fine-tuning. In this work, we propose Task-Aware Prompting and Selective Layer Fine-Tuning (TAP-SLF), a unified framework for multi-task ultrasound image analysis. TAP-SLF incorporates task-aware soft prompts to encode task-specific priors into the input token sequence and applies LoRA to selected specific top layers of the encoder. This strategy updates only a small fraction of the VFM parameters while keeping the pre-trained backbone frozen. By combining task-aware prompts with selective high-layer fine-tuning, TAP-SLF enables efficient VFM adaptation to diverse medical tasks within a shared backbone. Results on the FMC_UIA 2026 Challenge test set, where TAP-SLF wins fifth place, combined with evaluations on the officially released training dataset using an 8:2 train-test split, demonstrate that task-aware prompting and selective layer tuning are effective strategies for efficient VFM adaptation.

Paperid: 415, https://arxiv.org/pdf/2603.00431.pdf GitHub

Abstract:
A high-performing, general-purpose visual understanding model should map visual inputs to a taxonomic tree of labels, identify novel categories beyond the training set for which few or no publicly available images exist. Large Multimodal Models (LMMs) have achieved remarkable progress in fine-grained visual recognition (FGVR) for known categories. However, they remain limited in hierarchical visual recognition (HVR) that aims at predicting consistent label paths from coarse to fine categories, especially for novel categories. To tackle these challenges, we propose Taxonomy-Aware Representation Alignment (TARA), a simple yet effective strategy to inject taxonomic knowledge into LMMs. TARA leverages representations from biology foundation models (BFMs) that encode rich biological relationships through hierarchical contrastive learning. By aligning the intermediate representations of visual features with those of BFMs, LMMs are encouraged to extract discriminative visual cues well structured in the taxonomy tree. Additionally, we align the representations of the first answer token with the ground-truth label, flexibly bridging the gap between contextualized visual features and categories of varying granularity according to user intent. Experiments demonstrate that TARA consistently enhances LMMs' hierarchical consistency and leaf node accuracy, enabling reliable recognition of both known and novel categories within complex biological taxonomies. Code is available at https://github.com/PKU-ICST-MIPL/TARA_CVPR2026.

Paperid: 416, https://arxiv.org/pdf/2603.00413.pdf GitHub

Abstract:
Reconstructing transparent objects from a set of multi-view images is a challenging task due to the complicated nature and indeterminate behavior of light propagation. Typical methods are primarily tailored to specific scenarios, such as objects following a uniform topology, exhibiting ideal transparency and surface specular reflections, or with only surface materials, which substantially constrains their practical applicability in real-world settings. In this work, we propose a differentiable rendering framework for transparent objects, dubbed DiffTrans, which allows for efficient decomposition and reconstruction of the geometry and materials of transparent objects, thereby reconstructing transparent objects accurately in intricate scenes with diverse topology and complex texture. Specifically, we first utilize FlexiCubes with dilation and smoothness regularization as the iso-surface representation to reconstruct an initial geometry efficiently from the multi-view object silhouette. Meanwhile, we employ the environment light radiance field to recover the environment of the scene. Then we devise a recursive differentiable ray tracer to further optimize the geometry, index of refraction and absorption rate simultaneously in a unified and end-to-end manner, leading to high-quality reconstruction of transparent objects in intricate scenes. A prominent advantage of the designed ray tracer is that it can be implemented in CUDA, enabling a significantly reduced computational cost. Extensive experiments on multiple benchmarks demonstrate the superior reconstruction performance of our DiffTrans compared with other methods, especially in intricate scenes involving transparent objects with diverse topology and complex texture. The code is available at https://github.com/lcp29/DiffTrans.

Paperid: 417, https://arxiv.org/pdf/2603.00412.pdf GitHub GitHub

Abstract:
The development of 3D Vision-Language Models (VLMs), crucial for applications in robotics, autonomous driving, and augmented reality, is severely constrained by the scarcity of paired 3D-text data. Existing methods rely solely on next-token prediction loss, using only language tokens for supervision. This results in inefficient utilization of limited 3D data and leads to a significant degradation and loss of valuable geometric information in intermediate representations. To address these limitations, we propose {\mname}, a novel feature-level alignment regularization method. {\mname} explicitly supervises intermediate point cloud tokens to preserve fine-grained 3D geometric-semantic information throughout the language modeling process. Specifically, we constrain the intermediate point cloud tokens within the LLM to align with visual input tokens via a consistency loss. By training only a lightweight alignment projector and LoRA adapters, {\mname} achieves explicit feature-level supervision with minimal computational overhead, effectively preventing geometric degradation. Extensive experiments on ModelNet40 and Objaverse datasets demonstrate that our method achieves \textbf{2.08} pp improvement on average for classification tasks, with a substantial \textbf{7.50} pp gain on the challenging open-vocabulary Objaverse classification task and \textbf{4.88} pp improvement on 3D object captioning evaluated by Qwen2-72B-Instruct, validating the effectiveness of {\mname}. Code is publicly available at \href{https://github.com/yharoldsu0627/PointAlign}{https://github.com/yharoldsu0627/PointAlign}.

Paperid: 418, https://arxiv.org/pdf/2603.00349.pdf GitHub

Abstract:
Real-world scenarios increasingly require multiple embodied agents to collaborate in dynamic environments under embodied constraints, as many tasks exceed the capabilities of any single agent. Recent advances in large language models (LLMs) enable high-level cognitive coordination through reasoning, planning, and natural language communication. However, fine-grained analyses of how such collaboration emerges, unfolds, and contributes to task success in embodied multi-agent systems are difficult to conduct with existing benchmarks. In this paper, we introduce EmCoop, a benchmark framework for studying cooperation in LLM-based embodied multi-agent systems. Our framework separates a high-level cognitive layer from a low-level embodied interaction layer, allowing us to characterize agent cooperation through their interleaved dynamics over time. Given a cooperation-constrained embodied task, we propose generalizable, process-level metrics that diagnose collaboration quality and failure modes, beyond final task success. We instantiate our framework in two embodied environments that scale to arbitrary numbers of agents and support diverse communication topologies, and use these instantiations to demonstrate how EmCoop enables systematic analysis of cooperation dynamics across team sizes and task settings. The project web page can be found at: https://happyeureka.github.io/emcoop.

Paperid: 419, https://arxiv.org/pdf/2603.00337.pdf GitHub

Abstract:
Low-light images often suffer from low contrast, noise, and color distortion, degrading visual quality and impairing downstream vision tasks. We propose a novel conditional diffusion framework for low-light image enhancement that incorporates a Structured Control Embedding Module (SCEM). SCEM decomposes a low-light image into four informative components including illumination, illumination-invariant features, shadow priors, and color-invariant cues. These components serve as control signals that condition a U-Net-based diffusion model trained with a simplified noise-prediction loss. Thus, the proposed SCEM equipped Diffusion method enforces structured enhancement guided by physical priors. In experiments, our model is trained only on the LOLv1 dataset and evaluated without fine-tuning on LOLv2-real, LSRW, DICM, MEF, and LIME. The method achieves state-of-the-art performance in quantitative and perceptual metrics, demonstrating strong generalization across benchmarks. https://casted.github.io/scem/.

Paperid: 420, https://arxiv.org/pdf/2603.00309.pdf GitHub

Abstract:
The increasingly popular agentic AI paradigm promises to harness the power of multiple, general-purpose large language model (LLM) agents to collaboratively complete complex tasks. While many agentic AI systems utilize predefined workflows or agent roles in order to reduce complexity, ideally these agents would be truly autonomous, able to achieve emergent collaboration even as the number of collaborating agents increases. Yet in practice, such unstructured interactions can lead to redundant work and cascading failures that are difficult to interpret or correct. In this work, we study multi-agent systems composed of general-purpose LLM agents that operate without predefined roles, control flow, or communication constraints, relying instead on emergent collaboration to solve problems. We introduce the Dynamic Interaction Graph (DIG), which captures emergent collaboration as a time-evolving causal network of agent activations and interactions. DIG makes emergent collaboration observable and explainable for the first time, enabling real-time identification, explanation, and correction of collaboration-induced error patterns directly from agents' collaboration paths. Thus, DIG fills a critical gap in understanding how general LLM agents solve problems together in truly agentic multi-agent systems. The project webpage can be found at: https://happyeureka.github.io/dig.

Paperid: 421, https://arxiv.org/pdf/2603.00205.pdf GitHub

Abstract:
Generative models, particularly Diffusion Models (DM), have shown strong potential for Computed Tomography (CT) reconstruction serving as expressive priors for solving ill-posed inverse problems. However, diffusion-based reconstruction relies on Stochastic Differential Equations (SDEs) for forward diffusion and reverse denoising, where such stochasticity can interfere with repeated data consistency corrections in CT reconstruction. Since CT reconstruction is often time-critical in clinical and interventional scenarios, improving reconstruction efficiency is essential. In contrast, Flow Matching (FM) models sampling as a deterministic Ordinary Differential Equation (ODE), yielding smooth trajectories without stochastic noise injection. This deterministic formulation is naturally compatible with repeated data consistency operations. Furthermore, we observe that FM-predicted velocity fields exhibit strong correlations across adjacent steps. Motivated by this, we propose an FM-based CT reconstruction framework (FMCT) and an efficient variant (EFMCT) that reuses previously predicted velocity fields over consecutive steps to substantially reduce the number of Neural network Function Evaluations (NFEs), thereby improving inference efficiency. We provide theoretical analysis showing that the error introduced by velocity reuse is bounded when combined with data consistency operations. Extensive experiments demonstrate that FMCT/EFMCT achieve competitive reconstruction quality while significantly improving computational efficiency compared with diffusion-based methods. The codebase is open-sourced at https://github.com/EFMCT/EFMCT.

Paperid: 422, https://arxiv.org/pdf/2603.00195.pdf GitHub

Abstract:
The rapid proliferation of agentic AI skill ecosystems -- exemplified by OpenClaw (228,000 GitHub stars) and Anthropic Agent Skills (75,600 stars) -- has introduced a critical supply chain attack surface. The ClawHavoc campaign (January-February 2026) infiltrated over 1,200 malicious skills into the OpenClaw marketplace, while MalTool catalogued 6,487 malicious tools that evade conventional detection. In response, twelve reactive security tools emerged, yet all rely on heuristic methods that provide no formal guarantees. We present SkillFortify, the first formal analysis framework for agent skill supply chains, with six contributions: (1) the DY-Skill attacker model, a Dolev-Yao adaptation to the five-phase skill lifecycle with a maximality proof; (2) a sound static analysis framework grounded in abstract interpretation; (3) capability-based sandboxing with a confinement proof; (4) an Agent Dependency Graph with SAT-based resolution and lockfile semantics; (5) a trust score algebra with formal monotonicity; and (6) SkillFortifyBench, a 540-skill benchmark. SkillFortify achieves 96.95% F1 (95% CI: [95.1%, 98.4%]) with 100% precision and 0% false positive rate on 540 skills, while SAT-based resolution handles 1,000-node graphs in under 100 ms.

Paperid: 423, https://arxiv.org/pdf/2603.00187.pdf GitHub

Abstract:
To integrate seamlessly into real-world software engineering, Code Agents must evolve from passive instruction followers into proactive collaborative partners. However, current evaluation paradigms predominantly reward "guessing" user intent under ideal conditions, neglecting the agent's ability to align with users through dialogue--a critical trait for collaborative intelligence. In this work, we propose a paradigm shift in evaluation to drive this transition. We introduce ClarEval, a framework designed to assess an agent's "Collaborative Quotient" by simulating the inherent ambiguity of human communication. By systematically injecting three types of realistic ambiguity (missing goals, premises, and ambiguous terminology) into standard tasks, we force agents to step out of their "generator" role and engage in requirement elicitation. To quantify this capability, we propose a metric suite led by Average Turns to Clarify (ATC) and Key Question Coverage (KQC), which measure not just the correctness of the generated code, but the efficiency and precision of the collaboration. Our experiments on eleven state-of-the-art agents reveal a stark reality: while models like GPT-5-Coder excel at coding, they often lack the strategic communication skills required for efficient partnership. ClarEval thus serves as a crucial roadmap for bridging the gap between strong coders and capable collaborators.The code is available at https://github.com/JialinLi13/ClarEval

Paperid: 424, https://arxiv.org/pdf/2603.00164.pdf GitHub

Abstract:
We introduce Reverse CAPTCHA, an evaluation framework that tests whether large language models follow invisible Unicode-encoded instructions embedded in otherwise normal-looking text. Unlike traditional CAPTCHAs that distinguish humans from machines, our benchmark exploits a capability gap: models can perceive Unicode control characters that are invisible to human readers. We evaluate five models from two providers across two encoding schemes (zero-width binary and Unicode Tags), four hint levels, two payload framings, and with tool use enabled or disabled. Across 8,308 model outputs, we find that tool use dramatically amplifies compliance (Cohen's h up to 1.37, a large effect), that models exhibit provider-specific encoding preferences (OpenAI models decode zero-width binary; Anthropic models prefer Unicode Tags), and that explicit decoding instructions increase compliance by up to 95 percentage points within a single model and encoding. All pairwise model differences are statistically significant (p < 0.05, Bonferroni-corrected). These results highlight an underexplored attack surface for prompt injection via invisible Unicode payloads.

Paperid: 425, https://arxiv.org/pdf/2603.00155.pdf GitHub

Abstract:
Automated academic poster generation aims to distill lengthy research papers into concise, visually coherent presentations. Existing Multimodal Large Language Models (MLLMs) based approaches, however, suffer from three critical limitations: low information density in full-paper inputs, excessive token consumption, and unreliable layout verification. We present EfficientPosterGen, an end-to-end framework that addresses these challenges through semantic-aware retrieval and token-efficient multimodal generation. EfficientPosterGen introduces three core innovations: (1) Semantic-aware Key Information Retrieval (SKIR), which constructs a semantic contribution graph to model inter-segment relationships and selectively preserves important content; (2) Visual-based Context Compression (VCC), which renders selected text segments into images to shift textual information into the visual modality, significantly reducing token usage while generating poster-ready bullet points; and (3) Agentless Layout Violation Detection (ALVD), a deterministic color-gradient-based algorithm that reliably detects content overflow and spatial sparsity without auxiliary MLLMs. Extensive experiments demonstrate that EfficientPosterGen achieves substantial improvements in token efficiency and layout reliability while maintaining high poster quality, offering a scalable solution for automated academic poster generation. Our code is available at https://github.com/vinsontang1/EfficientPosterGen-Code.

Paperid: 426, https://arxiv.org/pdf/2603.00152.pdf GitHub

Abstract:
Following the success of Group Relative Policy Optimization (GRPO) in foundation LLMs, an increasing number of works have sought to adapt GRPO to Visual Large Language Models (VLLMs) for visual perception tasks (e.g., detection and segmentation). However, much of this line of research rests on a long-standing yet unexamined assumption: training paradigms developed for language reasoning can be transferred seamlessly to visual perception. Our experiments show that this assumption is not valid, revealing intrinsic differences between reasoning-oriented and perception-oriented settings. Using reasoning segmentation as a representative case, we surface two overlooked factors: (i) the need for a broader output space, and (ii) the importance of fine-grained, stable rewards. Building on these observations, we propose Dr.~Seg, a simple, plug-and-play GRPO-based framework consisting of a Look-to-Confirm mechanism and a Distribution-Ranked Reward module, requiring no architectural modifications and integrating seamlessly with existing GRPO-based VLLMs. Extensive experiments demonstrate that Dr.~Seg improves performance in complex visual scenarios while maintaining strong generalization. Code, models, and datasets are available at https://github.com/eVI-group-SCU/Dr-Seg.

Paperid: 427, https://arxiv.org/pdf/2603.00150.pdf GitHub

Abstract:
In this paper, we highlight a critical threat posed by emerging neural models: data plagiarism. We demonstrate how modern neural models (e.g., diffusion models) can replicate copyrighted images, even when protected by advanced watermarking techniques. To expose vulnerabilities in copyright protection and facilitate future research, we propose a general approach to neural plagiarism that can either forge replicas of copyrighted data or introduce copyright ambiguity. Our method, based on "anchors and shims", employs inverse latents as anchors and finds shim perturbations that gradually deviate the anchor latents, thereby evading watermark or copyright detection. By applying perturbations to the cross-attention mechanism at different timesteps, our approach induces varying degrees of semantic modification in copyrighted images, enabling it to bypass protections ranging from visible trademarks and signatures to invisible watermarks. Notably, our method is a purely gradient-based search that requires no additional training or fine-tuning. Experiments on MS-COCO and real-world copyrighted images show that diffusion models can replicate copyrighted images, underscoring the urgent need for countermeasures against neural plagiarism.

Paperid: 428, https://arxiv.org/pdf/2603.00149.pdf GitHub

Abstract:
Existing image SR and generic diffusion models transfer poorly to fluid SR: they are sampling-intensive, ignore physical constraints, and often yield spectral mismatch and spurious divergence. We address fluid super-resolution (SR) with \textbf{ReMD} (\underline{Re}sidual-\underline{M}ultigrid \underline{D}iffusion), a physics-consistent diffusion framework. At each reverse step, ReMD performs a \emph{multigrid residual correction}: the update direction is obtained by coupling data consistency with lightweight physics cues and then correcting the residual across scales; the multiscale hierarchy is instantiated with a \emph{multi-wavelet} basis to capture both large structures and fine vortical details. This coarse-to-fine design accelerates convergence and preserves fine structures while remaining equation-free. Across atmospheric and oceanic benchmarks, ReMD improves accuracy and spectral fidelity, reduces divergence, and reaches comparable quality with markedly fewer sampling steps than diffusion baselines. Our results show that enforcing physics consistency \emph{inside} the diffusion process via multigrid residual correction and multi-wavelet multiscale modeling is an effective route to efficient fluid SR. Our code are available on https://github.com/lizhihao2022/ReMD.

Paperid: 429, https://arxiv.org/pdf/2603.00143.pdf GitHub

Abstract:
Self-supervised vision models have achieved notable success in digital pathology. However, their domain-agnostic transformer architectures are not originally designed to account for fundamental biological elements of histopathology images, namely cells and their complex interactions. In this work, we hypothesize that a biologically-informed modeling of tissues as cell graphs offers a more efficient representation learning. Thus, we introduce GrapHist, a novel graph-based self-supervised learning framework for histopathology, which learns generalizable and structurally-informed embeddings that enable diverse downstream tasks. GrapHist integrates masked autoencoders and heterophilic graph neural networks that are explicitly designed to capture the heterogeneity of tumor microenvironments. We pre-train GrapHist on a large collection of 11 million cell graphs derived from breast tissues and evaluate its transferability across in- and out-of-domain benchmarks. Our results show that GrapHist achieves competitive performance compared to its vision-based counterparts in slide-, region-, and cell-level tasks, while requiring four times fewer parameters. It also drastically outperforms fully-supervised graph models on cancer subtyping tasks. Finally, we also release five graph-based digital pathology datasets used in our study at https://huggingface.co/ogutsevda/datasets , establishing the first large-scale graph benchmark in this field. Our code is available at https://github.com/ogutsevda/graphist .

Paperid: 430, https://arxiv.org/pdf/2603.00140.pdf GitHub

Abstract:
Text-to-image diffusion models often memorize training data, revealing a fundamental failure to generalize beyond the training set. Current mitigation strategies typically sacrifice image quality or prompt alignment to reduce memorization. To address this, we propose Reachability-Aware Diffusion Steering (RADS), an inference-time framework that prevents memorization while preserving generation fidelity. RADS models the diffusion denoising process as a dynamical system and applies concepts from reachability analysis to approximate the "backward reachable tube"--the set of intermediate states that inevitably evolve into memorized samples. We then formulate mitigation as a constrained reinforcement learning (RL) problem, where a policy learns to steer the trajectory away from memorization via minimal perturbations in the caption embedding space. Empirical evaluations show that RADS achieves a superior Pareto frontier between generation diversity (SSCD), quality (FID), and alignment (CLIP) compared to state-of-the-art baselines. Crucially, RADS provides robust mitigation without modifying the diffusion backbone, offering a plug-and-play solution for safe generation. Our website is available at: https://s-karnik.github.io/rads-memorization-project-page/.

Paperid: 431, https://arxiv.org/pdf/2603.00131.pdf GitHub

Abstract:
Subliminal prompting is a phenomenon in which language models are biased towards certain concepts or traits through prompting with semantically unrelated tokens. While prior work has examined subliminal prompting in user-LLM interactions, potential bias transfer in multi-agent systems and its associated security implications remain unexplored. In this work, we show that a single subliminally prompted agent can spread a weakening but persisting bias throughout its entire network. We measure this phenomenon across 6 agents using two different topologies, observing that the transferred concept maintains an elevated response rate throughout the network. To exemplify potential misalignment risks, we assess network performance on multiple-choice TruthfulQA, showing that subliminal prompting of a single agent may degrade the truthfulness of other agents. Our findings reveal that subliminal prompting introduces a new attack vector in multi-agent security, with implications for the alignment of such systems. The implementation of all experiments is publicly available at https://github.com/Multi-Agent-Security-Initiative/thought_virus .

Paperid: 432, https://arxiv.org/pdf/2603.00099.pdf GitHub

Abstract:
Neural architecture search (NAS) automates the discovery of neural networks that meet specified criteria, yet its evaluation procedures are often hardcoded, limiting the ability to introduce new metrics. This issue is especially pronounced in hardware-aware NAS, where objectives depend on target devices such as edge hardware. To address this limitation, we propose SEval-NAS, a metric-evaluation mechanism that converts architectures to strings, embeds them as vectors, and predicts performance metrics. Using NATS-Bench and HW-NAS-Bench, we evaluated accuracy, latency, and memory. Kendall's $τ$ correlations showed stronger latency and memory predictions than accuracy, indicating the suitability of SEval-NAS as a hardware cost predictor. We further integrated SEval-NAS into FreeREA to evaluate metrics not originally included. The method successfully ranked FreeREA-generated architectures, maintained search time, and required minimal algorithmic changes. Our implementation is available at: https://github.com/Analytics-Everywhere-Lab/neural-architecture-search

Paperid: 433, https://arxiv.org/pdf/2603.00097.pdf GitHub GitHub

Abstract:
Adverse Drug Reactions (ADRs) are a leading cause of morbidity and mortality. Existing prediction methods rely mainly on chemical similarity, machine learning on structured databases, or isolated target profiles, but often fail to integrate heterogeneous, partly unstructured evidence effectively. We present a knowledge graph-based framework that unifies diverse sources, drug-target data (ChEMBL), clinical trial literature (PubMed), trial metadata (ClinicalTrials.gov), and post-marketing safety reports (FAERS) into a single evidence-weighted bipartite network of drugs and medical conditions. Applied to 400 protein kinase inhibitors, the resulting network enables contextual comparison of efficacy (HR, PFS, OS), phenotypic and target similarity, and ADR prediction via target-to-adverse-event correlations. A non-small cell lung cancer case study correctly highlights established and candidate drugs, target communities (ERbB, ALK, VEGF), and tolerability differences. Designed as an orthogonal, extensible analysis and search tool rather than a replacement for current models, the framework excels at revealing complex patterns, supporting hypothesis generation, and enhancing pharmacovigilance. Code and data are publicly available at https://github.com/davidjackson99/PKI_KG.

Paperid: 434, https://arxiv.org/pdf/2603.00084.pdf GitHub

Abstract:
LLM-agents are increasingly used to accelerate the progress of scientific research. Yet a persistent bottleneck is data access: agents not only lack readily available tools for retrieval, but also have to work with unstrcutured, human-centric data on the Internet, such as HTML web-pages and PDF files, leading to excessive token consumption, limit working efficiency, and brittle evidence look-up. This gap motivates the development of \textit{an agentic data interface}, which is designed to enable agents to access and utilize scientific literature in a more effective, efficient, and cost-aware manner. In this paper, we introduce DeepXiv-SDK, which offers a three-layer agentic data interface for scientific literature. 1) Data Layer, which transforms unstructured, human-centric data into normalized and structured representations in JSON format, improving data usability and enabling progressive accessibility of the data. 2) Service Layer, which presents readily available tools for data access and ad-hoc retrieval. It also enables a rich form of agent usage, including CLI, MCP, and Python SDK. 3) Application Layer, which creates a built-in agent, packaging basic tools from the service layer to support complex data access demands. DeepXiv-SDK currently supports the complete ArXiv corpus, and is synchronized daily to incorporate new releases. It is designed to extend to all common open-access corpora, such as PubMed Central, bioRxiv, medRxiv, and chemRxiv. We release RESTful APIs, an open-source Python SDK, and a web demo showcasing deep search and deep research workflows. DeepXiv-SDK is free to use with registration.

Paperid: 435, https://arxiv.org/pdf/2603.00080.pdf GitHub

Abstract:
The exponential growth of scientific literature, datasets, and code repositories has created a discovery bottleneck that impedes knowledge synthesis and reproducibility. Traditional dissemination formats -- static PDFs, siloed code hosting, and fragmented data repositories -- fail to represent the interconnected narrative of modern research, while conventional metrics such as the H-index neglect contributions from reusable code and shared datasets. We present ResearchTwin, an open-source federated platform that transforms a researcher's scholarly output into a conversational digital twin, with a preliminary evaluation of its deployed prototype. The system uses a Bimodal Glial-Neural Optimization (BGNO) architecture comprising a Multi-Modal Connector Layer, a Glial Layer for caching and rate management, and a Neural Layer implementing Retrieval-Augmented Generation with a provider-agnostic LLM backend. We formalize the S-index, building on our earlier QIC framework, into a composite metric that extends FAIR principles -- via a binary accessibility/licensing gate, field-normalized impact scoring, and geometric collaboration scaling -- to quantify multimodal research impact. A case study comparing two researchers with similar H-indexes but substantially different S-indexes demonstrates that the metric captures dimensions of impact -- particularly dataset and code contributions -- invisible to citation-based measures alone. ResearchTwin exposes an inter-agentic discovery API using Schema.org typed responses and HATEOAS navigation, enabling AI agents to discover cross-lab synergies. A three-tier federated architecture preserves data sovereignty while enabling global discoverability.

Paperid: 436, https://arxiv.org/pdf/2603.00055.pdf GitHub

Abstract:
Although multimodal large language models (MLLMs) have advanced industrial anomaly detection toward a zero-shot paradigm, they still tend to produce high-confidence yet unreliable decisions in fine-grained and structurally complex industrial scenarios, and lack effective self-corrective mechanisms. To address this issue, we propose M3-AD, a unified reflection-aware multimodal framework for industrial anomaly detection. M3-AD comprises two complementary data resources: M3-AD-FT, designed for reflection-aligned fine-tuning, and M3-AD-Bench, designed for systematic cross-category evaluation, together providing a foundation for reflection-aware learning and reliability assessment. Building upon this foundation, we propose RA-Monitor, which models reflection as a learnable decision revision process and guides models to perform controlled self-correction when initial judgments are unreliable, thereby improving decision robustness. Extensive experiments conducted on M3-AD-Bench demonstrate that RA-Monitor outperforms multiple open-source and commercial MLLMs in zero-shot anomaly detection and anomaly analysis tasks. Code will be released at https://github.com/Yanhui-Lee/M3-AD.

Paperid: 437, https://arxiv.org/pdf/2603.00045.pdf GitHub

Abstract:
Diffusion language models theoretically allow for efficient parallel generation but are practically hindered by the "factorization barrier": the assumption that simultaneously predicted tokens are independent. This limitation forces a trade-off: models must either sacrifice speed by resolving dependencies sequentially or suffer from incoherence due to factorization. We argue that this barrier arises not from limited backbone expressivity, but from a structural misspecification: models are restricted to fully factorized outputs because explicitly parameterizing a joint distribution would require the Transformer to output a prohibitively large number of parameters. We propose Coupled Discrete Diffusion (CoDD), a hybrid framework that breaks this barrier by replacing the fully-factorized output distribution with a lightweight, tractable probabilistic inference layer. This formulation yields a distribution family that is significantly more expressive than standard factorized priors, enabling the modeling of complex joint dependencies, yet remains compact enough to avoid the prohibitive parameter explosion associated with full joint modeling. Empirically, CoDD seamlessly enhances diverse diffusion language model architectures with negligible overhead, matching the reasoning performance of computationally intensive Reinforcement Learning baselines at a fraction of the training cost. Furthermore, it prevents performance collapse in few-step generation, enabling high-quality outputs at significantly reduced latencies. Code available at: https://github.com/liuanji/CoDD

Paperid: 438, https://arxiv.org/pdf/2603.00039.pdf GitHub GitHub

Abstract:
LLM-as-a-judge ensembles are the standard paradigm for scalable evaluation, but their aggregation mechanisms suffer from a fundamental flaw: they implicitly assume that judges provide independent estimates of true quality. However, in practice, LLM judges exhibit correlated errors caused by shared latent confounders -- such as verbosity, stylistic preferences, or training artifacts -- causing standard aggregation rules like majority vote or averaging to provide little gain or even amplify systematic mistakes. To address this, we introduce CARE, a confounder-aware aggregation framework that explicitly models LLM judge scores as arising from both a latent true-quality signal and shared confounding factors. Rather than heuristically re-weighting judges, CARE separates quality from confounders without access to ground-truth labels. We provide theoretical guarantees for identifiability and finite-sample recovery under shared confounders, and we quantify the systematic bias incurred when aggregation models omit confounding latent factors. Across 12 public benchmarks spanning continuous scoring, binary classification, and pairwise preference settings, CARE improves aggregation accuracy, reducing error by up to 26.8\%. Code is released in \href{https://github.com/SprocketLab/CARE}{https://github.com/SprocketLab/CARE}.