AI papers — 2026-10-14

Research this morning focused on several areas, including predicting where RNA molecules end up inside a cell to understand gene function. A key piece of work involves using diffusion models to shrink proteins while keeping their function intact, as described in A Diffusion Model to Shrink Proteins While Maintaining Their Function, which addresses the physical constraints of molecular interactions.

Another area explored is how spatial structure and information fidelity affect gene expression-to-image learning, which Understanding Gene Expression-to-Image Learning: Spatial Structure and Information Fidelity investigates. This work helps clarify the link between a cell's physical layout and how its genes are expressed.

Researchers also used MRFlow to predict virtual cell perturbations by employing mechanism-structured residual flow matching, as detailed in MRFlow: Mechanism-Structured Residual Flow Matching for Virtual Cell Perturbation Prediction. This technique allows simulations of changes in cell structure to see the resulting biological system changes.

Finally, agentic program search was used to discover transferable nucleic acid design algorithms with AutoNovo AutoNovo: Agentic Program Search Discovers Transferable Nucleic Acid Design Algorithms. This suggests new methods for designing genetic sequences based on learned principles.

The most significant development today is the work on Graph Neural Network-Based Reinforcement Learning for Controlling Biological Networks, which deals with managing complex biological systems using learning techniques. This research introduces the GATTACA Framework to control biological networks by using graph neural networks to guide reinforcement learning decisions.

This approach builds upon earlier work that looked at controlling biological networks, specifically through the GATTACA Framework GATTACA, which uses a graph neural network structure for this purpose. Another piece of research provides a deeper look into modeling neural activity using Deep probabilistic model synthesis, which aims to create unified models of whole-brain activity across different subjects Deep probabilistic model synthesis.

In terms of perception and learning, there is work on learning perceptual geometry from natural odor mixtures using LLM-derived descriptors Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors. This shows how large language models can help agents learn to perceive complex sensory information. This is related to the idea of using learned representations in control systems, as seen in MFG: Multi-turn Feedback Guided Molecular Optimization <a href="/papers/2610.18851</a>, where feedback guides molecular optimization steps.

Studies were also conducted on how to diagnose opponent modeling within large language models Beyond Successful Play: Diagnosing Opponent Modeling in Large Language Models and how neuroeconomics can inform policy decisions Neural Evidence and Behavioural Welfare Economics: When Can Neuroeconomics Inform Policy?, showing the breadth of current research across AI, neuroscience, and economics.

The most significant development today involves understanding how the brain prioritizes internal states for theory of mind reasoning, explored in Learning Dynamic Belief Graphs for Theory-of-mind Reasoning. This work attempts to model how belief structures evolve over time by learning dynamic graphs that represent these beliefs.

This modeling effort builds upon earlier work that focused on spatial domain identification, specifically MP-MNCA, which is designed for efficient spatial domain identification using morphology and masked neighbor cross-attention. This suggests a connection between mapping complex spatial relationships and tracking evolving mental states.

Another important piece is the attempt to learn how to optimize internal prioritization by dissociating likelihood precision from preference reweighting in Active Inference, detailed in Two Routes to Interoceptive Prioritisation: Dissociating Likelihood Precision and Preference Reweighting in Active Inference. This framework tries to separate how accurately we estimate a belief from how much we value that belief, which is a key step toward understanding interoceptive prioritization.

This concept of dynamic belief graphs seems related to how different systems might learn representations, similar to what is explored in Semantic Context Forests for Learning-Based Knee Cartilage Segmentation in 3D MR Images. This segmentation work uses context forests to learn from semantic information within medical images, suggesting a shared methodology for extracting meaningful patterns from complex data.

Finally, there is research on predicting stress-strain behaviors of additively manufactured materials via loss-based and activation-based physics-informed machine learning, which uses ML to predict material properties based on physical constraints. This connects to the broader theme of using learned models—whether for belief tracking or material science—to make predictions about complex systems.

The most significant piece of work today is the development of AeroJEPA, which tackles the challenge of learning semantic latent representations for scalable three-dimensional aerodynamic field modeling. This matters because accurately modeling these fields is crucial for designing efficient and robust aircraft systems.

This approach involves using transformers with physics-informed encodings to achieve robust detection of eccentric binary black holes in pulsar timing array data, as described in Transformers with Physics-Informed Encodings and Simulation-Based Inference for Robust Detection of Eccentric Binary Black Holes in Pulsar Timing Array Data. This work is important because it shows how to use simulation-based inference within transformer architectures for complex astrophysical data analysis.

Another development is QwenGyre, which presents an elastic reinforcement learning framework designed to train xLong-Horizon Agents. This framework is relevant because training agents over long time horizons requires a flexible learning mechanism that can adapt effectively during the process.

A decomposition-based approach for evaluating and analyzing inter-annotator disagreement was also seen, detailed in A Decomposition-Based Approach for Evaluating and Analyzing Inter-Annotator Disagreement. This method is useful because it provides a structured way to understand where human labeling errors occur when multiple people annotate the same data.

Agent-OM leverages large language model agents for ontology matching, which is relevant because matching ontologies using LLM agents can streamline complex knowledge integration tasks across different domains.

The most significant development today involves a robust autoencoder ensemble approach for anomaly detection in text, which offers a solid method for spotting unusual patterns in language. This work, AKMPC, tried combining multiple autoencoders to improve the accuracy of finding anomalies.

Another key piece of research focuses on building multilingual datasets for predicting mental health severity through large language models, which addresses the challenge of accessibility across different languages. This effort explores the prospects and challenges involved in using LLMs for this purpose.

There is also work on subgraphs-aware training of language models for knowledge graph completion, which is significant because it tries to incorporate structural information into how these models learn. This approach uses structure-aware contrastive learning to guide the model's training.

Following that, PEDRO presents a parameter-efficient fine-tuning method using prompt dependent representation modification. This technique aims to make fine-tuning large models more efficient by modifying the representation based on the prompt used during tuning.

The work on polyrating offers a cost-effective and bias-aware rating system for evaluating large language models, which is important because it provides a better way to assess model performance beyond simple metrics. This system helps in getting a more nuanced view of what LLMs are actually doing.

Finally, there is the unified approach to routing and cascading for LLMs, which deals with how to best direct and chain multiple large language models together for complex tasks.

The most significant development this morning involves understanding the emotional landscape of human communication, specifically how emotions shift within comments and how those shifts relate to responses generated by large language models. This is crucial because it informs how we might build more nuanced and empathetic AI systems.

Some work on analyzing emotional variances in human comments and LLM-generated responses with EmoXpt EmoXpt attempts to map these emotional shifts. This approach is important because it moves beyond simple sentiment analysis to capture the actual variation in feeling during interaction.

Another key area is tackling the problem of hallucination in large vision-language models through adaptive attention calibration, explored in Mitigating Hallucination in Large Vision-Language Models via Adaptive Attention Calibration <a href="/papers/2505.21472</a>. This work is important because it directly addresses the reliability issues inherent in these models when dealing with complex visual information.

Then there is the effort to improve reasoning within language models by incorporating multidimensional consistency, detailed in Multidimensional Consistency Improves Reasoning in Language Models <a href="/papers/2503.02670</a>. This helps make the model's internal logic more robust and consistent across different aspects of a problem.

Finally, there is research into how toxicity manifests as a property of communicative events through a contextual stress measurement framework for NLP <a href="/papers/2503.16072</a>. This provides a way to measure stress in communication, which is relevant when considering the emotional context explored by EmoXpt.

The most significant development today involves understanding how to properly evaluate large language models because without good benchmarks, we cannot reliably measure their true capabilities. Benchmarks Are Not Atomic: Composition-Aware LLM Evaluation using BenchHub addresses this by proposing a method for evaluating models based on the composition of their components rather than treating individual tests in isolation. This approach is crucial because it moves beyond simple pass or fail metrics to capture how different parts of a model interact and contribute to overall performance.

This idea connects with FailureSensorIQ, which creates a multi-choice question dataset designed to help us understand the relationships between sensors and potential failure modes, providing a structured way to probe system understanding. Furthermore, simulating student-tutor interactions with LLMs for multilingual math feedback offers a practical application of this kind of modeling by testing how these models handle complex pedagogical scenarios Simulating Student-Tutor Interactions with LLMs for Multilingual Math Feedback.

A more granular look at model behavior is provided by RCStat, which introduces a statistical framework for using relative contextualization within transformers to better understand how context influences the model's output RCStat: A Statistical Framework for using Relative Contextualization in Transformers. This statistical rigor complements the compositional evaluation approach of Benchmarks Are Not Atomic, suggesting that we need both structural and contextual insights to truly gauge LLM proficiency.

The most significant development today involves understanding how gender bias manifests in large language models through the lens of social relationships, which points toward deeper societal implications for AI deployment. This was explored in the work titled From Individuals to Interactions: Benchmarking Gender Bias in Multimodal Large Language Models from the Lens of Social Relationship AKMPC.

This research attempts to benchmark gender bias by examining interactions between individuals within multimodal LLMs, suggesting that simply looking at individual model outputs is insufficient to capture systemic bias. A more granular approach was taken with SciRerankBench, which benchmarks rerankers for scientific retrieval-augmented generation with LLMs SciRerankBench. This helps assess how well these systems prioritize relevant scientific information when using LLMs to generate answers.

Ongoing work is also being done into making agent systems more coherent through latent profiling, as seen in Are LLM Agents Behaviorally Coherent? A Latent Profiling Approach for Social Simulation <a href="/papers/2509.03736</a>. This method profiles the latent states of agents to see if their simulated social behaviors are consistent, which is important for building reliable multi-agent systems.

On a more technical level, efforts are being made to improve model efficiency and specialization, such as training Matryoshka Mixture-of-Experts for Elastic Inference-Time Expert Utilization <a href="/papers/2509.26520</a>. This technique allows the model to dynamically use only the necessary experts during inference, which is a key step toward practical deployment.

Finally, there are specific improvements in code generation and language modeling; IFEvalCode focuses on controlled code generation <a href="/papers/2507.22462</a>, while GHTM presents a graph-based hybrid topic modeling approach for Bengali language <a href="/papers/2508.00605</a>. These contributions represent the breadth of research across these different areas today.

Today's papers

• A Comprehensive Review on RNA Subcellular Localization Prediction. [paper]
• Understanding Gene Expression-to-Image Learning: Spatial Structure and Information Fidelity. [paper]
• A Diffusion Model to Shrink Proteins While Maintaining Their Function. [paper]
• MRFlow: Mechanism-Structured Residual Flow Matching for Virtual Cell Perturbation Prediction. [paper]
• AutoNovo: Agentic Program Search Discovers Transferable Nucleic Acid Design Algorithms. [paper]
• From Simulated Fluids to Living Tissue: Zero-Shot Foundation Model Transfer to Perturbed Embryos. [paper]
• MedEvo: Medical Reasoning Evolution through Knowledge Navigation. [paper]
• ECG-RSI: Recursive Self-Improvement through Targeted Electrocardiogram Waveform Evolution. [paper]
• MFG: Multi-turn Feedback Guided Molecular Optimization. [paper]
• Graph Neural Network-Based Reinforcement Learning for Controlling Biological Networks - the GATTACA Framework. [paper]
• Neural Evidence and Behavioural Welfare Economics: When Can Neuroeconomics Inform Policy?. [paper]
• NeuroVolve: Evolving Visual Stimuli toward Programmable Neural Objectives. [paper]
• Deep probabilistic model synthesis enables unified modeling of whole-brain neural activity across individual subjects. [paper]
• Interoceptive Attention as Dynamic Homeostatic Prioritization in a Foraging Agent. [paper]
• Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors. [paper]
• Beyond Successful Play: Diagnosing Opponent Modeling in Large Language Models. [paper]
• Two Routes to Interoceptive Prioritisation: Dissociating Likelihood Precision and Preference Reweighting in Active Inference. [paper]
• Semantic Context Forests for Learning-Based Knee Cartilage Segmentation in 3D MR Images. [paper]
• MP-MNCA: Morphology-Prior Masked Neighbor Cross-Attention for Efficient Spatial Domain Identification. [paper]
• Learning Dynamic Belief Graphs for Theory-of-mind Reasoning. [paper]
• TRACE: Expert-Aligned ECG Representation Learning with Rigorous Benchmarking and Real-World Validation in Acute Cardiac Care. [paper]
• A Differentiable Optimization Framework for Registering Sequential Bounding Boxes with Point Cloud Stream. [paper]
• SlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving. [paper]
• Predicting Stress-strain Behaviors of Additively Manufactured Materials via Loss-based and Activation-based Physics-informed Machine Learning. [paper]
• AeroJEPA: Learning Semantic Latent Representations for Scalable 3D Aerodynamic Field Modeling. [paper]
• Transformers with Physics-Informed Encodings and Simulation-Based Inference for Robust Detection of Eccentric Binary Black Holes in Pulsar Timing Array Data. [paper]
• QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents. [paper]
• A Decomposition-Based Approach for Evaluating and Analyzing Inter-Annotator Disagreement. [paper]
• Controlled Text Generation via Language Model Arithmetic. [paper]
• Agent-OM: Leveraging LLM Agents for Ontology Matching. [paper]
• Prediction Is All MoE Needs: Expert Load Distribution Goes from Fluctuating to Stabilizing. [paper]
• Identification of Entailment and Contradiction Relations for Natural Language Argumentation: A Neurosymbolic Approach. [paper]
• A Robust Autoencoder Ensemble-Based Approach for Anomaly Detection in Text. [paper]
• Subgraph-Aware Training of Language Models for Knowledge Graph Completion Using Structure-Aware Contrastive Learning. [paper]
• Polyrating: A Cost-Effective and Bias-Aware Rating System for LLM Evaluation. [paper]
• Building Multilingual Datasets for Predicting Mental Health Severity through LLMs: Prospects and Challenges. [paper]
• PEDRO: Parameter-Efficient Fine-tuning with Prompt DEpenDent Representation MOdification. [paper]
• Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise. [paper]
• A Unified Approach to Routing and Cascading for LLMs. [paper]
• Leveraging LLMs for Translating and Classifying Mental Health Data. [paper]
• EmoXpt: Analyzing Emotional Variances in Human Comments and LLM-Generated Responses. [paper]
• Open Foundation Models in Healthcare: Challenges, Paradoxes, and Opportunities with GenAI Driven Personalized Prescription. [paper]
• Natural Language Generation from Visual Events: State-of-the-Art and Key Open Questions. [paper]
• Multidimensional Consistency Improves Reasoning in Language Models. [paper]
• VQEL: Enabling Self-Play in Emergent Language Games via Agent-Internal Vector Quantization. [paper]
• Toxicity Is a Property of Communicative Events: A Contextual Stress Measurement Framework for NLP. [paper]
• Synergizing Large Language Models and Theory for Human-like Causal Reasoning. [paper]
• Mitigating Hallucination in Large Vision-Language Models via Adaptive Attention Calibration. [paper]
• Localizing Persona Representations in LLMs. [paper]
• Benchmarks Are Not Atomic: Composition-Aware LLM Evaluation using BenchHub. [paper]
• FailureSensorIQ: A Multi-Choice QA Dataset for Understanding Sensor Relationships and Failure Modes. [paper]
• Simulating Student-Tutor Interactions with LLMs for Multilingual Math Feedback. [paper]
• GUARD: Guided Unlearning and Retention via Data Attribution for Large Language Models. [paper]
• Cite Pretrain: Retrieval-Free Knowledge Attribution for Large Language Models. [paper]
• End-to-End Spoken Grammatical Error Correction. [paper]
• RCStat: A Statistical Framework for using Relative Contextualization in Transformers. [paper]
• From Individuals to Interactions: Benchmarking Gender Bias in Multimodal Large Language Models from the Lens of Social Relationship. [paper]
• IFEvalCode: Controlled Code Generation. [paper]
• GHTM: A graph-based hybrid topic modeling approach for Bengali language. [paper]
• RCR-Router: Efficient Role-Aware Context Routing for Multi-Agent LLM Systems with Structured Memory. [paper]

The papers

• Semantic Context Forests for Learning-Based Knee Cartilage Segmentation in 3D MR Images —
• Neural Ordinary Differential Equations —
• Bayesian Nonparametric Thompson Sampling for Multi-Armed Bandits —
• GEN: Highly Efficient SMILES Explorer Using Autodidactic Generative Examination Networks —
• SynDistNet: Self-Supervised Monocular Fisheye Camera Distance Estimation Synergized with Semantic Segmentation for Autonomous Driving —
• Transparency, Auditability and eXplainability of Machine Learning Models in Credit Scoring —
• The Hessian Screening Rule —
• NRC-GAMMA: Introducing a Novel Large Gas Meter Image Dataset —
• A Decomposition-Based Approach for Evaluating and Analyzing Inter-Annotator Disagreement —
• Benchopt: Reproducible, efficient and collaborative optimization benchmarks —
• Controlled Text Generation via Language Model Arithmetic —
• Agent-OM: Leveraging LLM Agents for Ontology Matching —
• Optimal Transport Aggregation for Distributed Mixture-of-Experts —
• Forest Inspection Dataset for Aerial Semantic Segmentation and Depth Estimation —
• Prediction Is All MoE Needs: Expert Load Distribution Goes from Fluctuating to Stabilizing —
• Identification of Entailment and Contradiction Relations for Natural Language Argumentation: A Neurosymbolic Approach —
• A Robust Autoencoder Ensemble-Based Approach for Anomaly Detection in Text —
• RDM: Recurrent Diffusion Model for Human Motion Generation —
• Subgraph-Aware Training of Language Models for Knowledge Graph Completion Using Structure-Aware Contrastive Learning —
• Polyrating: A Cost-Effective and Bias-Aware Rating System for LLM Evaluation —
• TropNNC: Structured Neural Network Compression Using Tropical Geometry —
• Building Multilingual Datasets for Predicting Mental Health Severity through LLMs: Prospects and Challenges —
• PEDRO: Parameter-Efficient Fine-tuning with Prompt DEpenDent Representation MOdification —
• Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise —
• A Unified Approach to Routing and Cascading for LLMs —
• Leveraging LLMs for Translating and Classifying Mental Health Data —
• Frequency-Guided Diffusion Model with Perturbation Training for Skeleton-Based Video Anomaly Detection —
• The Choice of Normalization Influences Shrinkage in Regularized Regression —
• EmoXpt: Analyzing Emotional Variances in Human Comments and LLM-Generated Responses —
• Open Foundation Models in Healthcare: Challenges, Paradoxes, and Opportunities with GenAI Driven Personalized Prescription —
• Differentially Private Synthetic Data via APIs 3: Using Simulators Instead of Foundation Model —
• Natural Language Generation from Visual Events: State-of-the-Art and Key Open Questions —
• Robustness and Cybersecurity in the EU Artificial Intelligence Act —
• Multidimensional Consistency Improves Reasoning in Language Models —
• VQEL: Enabling Self-Play in Emergent Language Games via Agent-Internal Vector Quantization —
• Universal Incremental Learning: Mitigating Confusion from Inter- and Intra-task Distribution Randomness —
• A Closer Look at Dynamic Scene Graph Generation In the Era of Multimodal Large Language Models —
• Toxicity Is a Property of Communicative Events: A Contextual Stress Measurement Framework for NLP —
• A Comprehensive Review on RNA Subcellular Localization Prediction —
• Graph Neural Network-Based Reinforcement Learning for Controlling Biological Networks - the GATTACA Framework —
• Panoramic Out-of-Distribution Segmentation —
• Towards Accurate State Estimation: Motion Dynamics Kalman Filter for 3D Multi-Object Tracking —
• Synergizing Large Language Models and Theory for Human-like Causal Reasoning —
• Towards Understanding Deep Learning Model in Image Recognition via Coverage Test —
• Facial Recognition Leveraging Generative Adversarial Networks —
• Generative AI for Autonomous Driving: A Review —
• Mitigating Hallucination in Large Vision-Language Models via Adaptive Attention Calibration —
• Localizing Persona Representations in LLMs —
• Benchmarks Are Not Atomic: Composition-Aware LLM Evaluation using BenchHub —
• FailureSensorIQ: A Multi-Choice QA Dataset for Understanding Sensor Relationships and Failure Modes —
• Simulating Student-Tutor Interactions with LLMs for Multilingual Math Feedback —
• GUARD: Guided Unlearning and Retention via Data Attribution for Large Language Models —
• Teleoperated Driving: a New Challenge for 3D Object Detection in Compressed Point Clouds —
• Cite Pretrain: Retrieval-Free Knowledge Attribution for Large Language Models —
• End-to-End Spoken Grammatical Error Correction —
• RCStat: A Statistical Framework for using Relative Contextualization in Transformers —
• From Individuals to Interactions: Benchmarking Gender Bias in Multimodal Large Language Models from the Lens of Social Relationship —
• The Serial Scaling Hypothesis —
• Missing value imputation with adversarial random forests -- MissARF —
• IFEvalCode: Controlled Code Generation —
• GHTM: A graph-based hybrid topic modeling approach for Bengali language —
• RCR-Router: Efficient Role-Aware Context Routing for Multi-Agent LLM Systems with Structured Memory —
• High-Order Error Bounds for Markovian LSA with Richardson-Romberg Extrapolation —
• One Size Does Not Fit All: Distribution-Guided Tensor-Wise Adaptive Sparsification for Model Merging —
• SciRerankBench: Benchmarking Rerankers Towards Scientific Retrieval-Augmented Generation with LLMs —
• Novel View Synthesis using DDIM Inversion —
• LoRAtorio: An intrinsic approach to LoRA Skill Composition —
• MMG-Vid: Maximizing Marginal Gains at Segment-level and Token-level for Efficient Video LLMs —
• Bouncy particle sampler with infinite exchanging parallel tempering —
• Are LLM Agents Behaviorally Coherent? A Latent Profiling Approach for Social Simulation —
• PanoLAM: Large Avatar Model for Gaussian Full-Head Synthesis from One-shot Unposed Image —
• Streamline pathology foundation model by cross-magnification distillation —
• Training Matryoshka Mixture-of-Experts for Elastic Inference-Time Expert Utilization —
• Computational Social Linguistics for Telugu Cultural Preservation: Novel Algorithms for Chandassu Metrical Pattern Recognition —
• Efficiently Generating Correlated Sample Paths from Multi-step Time Series Foundation Models —
• Frequency-Aware Model Parameter Explorer: A new attribution method for improving explainability —
• Sri Lanka Document Datasets: A Large-Scale, Multilingual Resource for Law, News, and Policy —
• Rethinking Langevin Thompson Sampling from A Stochastic Approximation Perspective —
• Bayesian Optimization under Uncertainty for Training a Scale Parameter in Stochastic Models —
• Contextual Representation Flow Across Depth in State-Space and Transformer Architectures —
• Hybrid Models for Natural Language Reasoning: The Case of Syllogistic Logic —
• Batch Bayesian Active Learning with Partial Batch Label Sampling —
• Fine-grained Analysis of Brain-LLM Alignment through Input Attribution —
• A Function-Centric Perspective on Flat and Sharp Minima —
• Text Anomaly Detection with Simplified Isolation Kernel —
• Mode Collapse of Mean-Field Variational Inference —
• Stochastic Difference-of-Convex Optimization with Momentum —
• OmniRaindrop: A Comprehensive Benchmark for Dual-Focused Raindrop Removal under Day and Night Scenes —
• An Explainable Hybrid AI Framework for Enhanced Tuberculosis and Symptom Detection —
• Deep Jump Gaussian Processes for Surrogate Modeling of High-Dimensional Piecewise Continuous Functions —
• Complexity Dependent Error Rates for Physics-informed Statistical Learning via the Small-ball Method —
• Iterative Critique-Refine Framework for Enhancing LLM Personalization —
• Scaling Latent Reasoning via Looped Language Models —
• Training LLMs Beyond Next Token Prediction - Filling the Mutual Information Gap —
• The Optimization Landscape of Carath'eodory Decomposition of Toeplitz Covariances —
• S 2LM: Towards Semantic Steganography via Large Language Models —
• A Diffusion Model to Shrink Proteins While Maintaining Their Function —
• Exploring the Underwater World Segmentation without Extra Training —
• The Impact of Longitudinal Mammogram Alignment on Breast Cancer Risk Assessment —
• Multi-agent Undercover Gaming: Hallucination Removal via Counterfactual Test for Multimodal Reasoning —
• From Classification to Cross-Modal Understanding: Leveraging Vision-Language Models for Fine-Grained Renal Pathology —
• D squared-VPR: A Parameter-efficient Visual-foundation-model-based Visual Place Recognition Method via Knowledge Distillation and Deformable Aggregation —
• Naga: Vedic Encoding for Deep State Space Models —
• UnSAMv2: Self-Supervised Learning Enables Segment Anything at Any Granularity —
• InstantViR: Real-Time Video Inverse Problem Solver with Distilled Diffusion Prior —
• ManipShield: A Unified Framework for Image Manipulation Detection, Localization and Explanation —
• Q-Save: Towards Scoring and Attribution for Generated Video Evaluation —
• Neural Evidence and Behavioural Welfare Economics: When Can Neuroeconomics Inform Policy? —
• Explainable Visual Anomaly Detection via Concept Bottleneck Models —
• Semantics as a Shield: Label Disguise Defense (LDD) against Prompt Injection in LLM Sentiment Classification —
• From Topology to Retrieval: Decoding Embedding Spaces with Unified Signatures —
• NeuroVolve: Evolving Visual Stimuli toward Programmable Neural Objectives —
• ELVIS: Enhance Low-Light for Video Instance Segmentation in the Dark —
• Control Consistency Losses for Diffusion Bridges —
• CHIMERA: Adaptive Cache Injection and Semantic Anchor Prompting for Zero-shot Image Morphing with Morphing-oriented Metrics —
• An Agentic AI System for Multi-Framework Communication Coding —
• Simpler and Stronger Baselines for All-in-One Image Restoration —
• AdvJudge-Zero: Binary Decision Flips in LLM-as-a-Judge via Adversarial Control Tokens —
• Sampling from multimodal distributions with warm starts: Non-asymptotic bounds for the Reweighted Annealed Leap-Point Sampler —
• KerJEPA: Kernel Discrepancies for Euclidean Self-Supervised Learning —
• Measuring all the noises of LLM Evals —
• Toward Zero-Shot Point Cloud Registration Across Diverse Scales, Scenes, and Sensor Setups —
• Mechanistic Knobs in LLMs: Retrieving and Steering High-Order Semantic Features via Sparse Autoencoders —
• SOVABench: A Vehicle Surveillance Action Retrieval Benchmark for Multimodal Large Language Models —
• Context-Aware Decoding for Faithful Vision-Language Generation —
• Nonparametric Kernel Clustering with Bandit Feedback —
• GeoMotionGPT: Geometry-Aligned Motion Understanding with Large Language Models —
• What Are We Measuring in NLG? A Meta-Analysis of Evaluation Trends 2020-2025 —
• Representational Similarity Alignment of Figurative Language in Humans and Large Language Models —
• Less is More: Label-Guided Efficient Summarization of Procedural and Instructional Videos —
• QASA: Quality-Guided K-Adaptive Slot Attention for Unsupervised Object-Centric Learning —
• Probe and Skip: Self-Predictive Token Skipping for Efficient Long-Context LLM Inference —
• Towards Token-Level Text Anomaly Detection —
• AfriEconQA: A Benchmark for Quantitative and Temporal Reasoning over World Bank Economic Reports —
• Where is the multimodal goal post? On the Ability of Foundation Models to Recognize Highlights in Videos —
• Retell, Reward, Repeat: Reinforcement Learning for Narrative Theory-Informed Story Retelling —
• Uni-RS: A Spatially Faithful Unified Understanding and Generation Model for Remote Sensing —
• PixSearch: Region-Grounded Retrieval for Knowledgeable Large Multimodal Models —
• The pretraining domain outweighs the training objective in setting the privacy-utility trade-off of differentially private medical image analysis —
• VERGE: Formal Refinement and Guidance Engine for Verifiable LLM Reasoning —
• TPGDiff: Hierarchical Triple-Prior Guided Diffusion for Image Restoration —
• GDCNet: Generative Discrepancy Comparison Network for Multimodal Sarcasm Detection —
• Signal from Structure: Exploiting Submodular Upper Bounds in Generative Flow Networks —
• BookNet: Book Image Rectification via Cross-Page Attention Network —
• OpenVTON-Bench: A Large-Scale High-Resolution Benchmark for Controllable Virtual Try-On Evaluation —
• From Associations to Activations: Comparing Behavioral and Hidden-State Semantic Geometry in LLMs —
• Fast Non-Episodic Finite-Horizon RL with K-Step Lookahead Thresholding —
• Distill3R: A Pipeline for Democratizing 3D Foundation Models on Commodity Hardware —
• DCD: Decomposition-based Causal Discovery from Autocorrelated and Non-Stationary Temporal Data —
• DRMOT: A Dataset and Framework for RGBD Referring Multi-Object Tracking —
• VRIQ: Benchmarking and Diagnosing the Visual-Reasoning IQ of Vision-Language Models —
• Causal Front-Door Adjustment for Robust Jailbreak Attacks on LLMs —
• LongR: Unleashing Long-Context Reasoning via Reinforcement Learning with Dense Utility Rewards —
• Equilibrium contrastive learning for imbalanced image classification —
• 3DXTalker: Unifying Identity, Lip Sync, Emotion, and Spatial Dynamics in Expressive 3D Talking Avatars —
• Learning Ordinal Probabilistic Reward from Preferences —
• Certified Per-Instance Unlearning Using Individual Sensitivity Bounds —
• VETime: Vision Enhanced Zero-Shot Time Series Anomaly Detection —
• A Multi-modal Detection System for Infrastructure-based Freight Signal Priority —
• HONEST-CAV: Hierarchical Optimization of Network Signals and Trajectories for Connected and Automated Vehicles with Multi-Agent Reinforcement Learning —
• TurkicNLP: An NLP Toolkit for Turkic Languages —
• A text-guided vision model for enhanced recognition of small instances —
• Rethinking Floorplan Generation with Space Syntax: From Design Theory to Model Improvement —
• Strategy Executability in Mathematical Reasoning: Leveraging Human-Model Differences for Effective Guidance —
• Toward Guarantees for Clinical Reasoning in Vision Language Models via Formal Verification —
• DreamWorld: Unified World Modeling in Video Generation —
• Tri-path DINO: Feature Complementary Learning for Remote Sensing Multi-Class Change Detection —
• TopoCurate:Modeling Interaction Topology for Tool-Use Agent Training —
• Learning Shortest Paths with Generative Flow Networks —
• Skin Tone in the Wild: A Dataset and Benchmark for 10-tone MST Classification —
• A Counterfactual Graph Approach to Oppositional Narrative Analysis —
• FusionRegister: Every Infrared and Visible Image Fusion Deserves Registration —
• HybridStitch: Pixel and Timestep Level Model Stitching for Diffusion Acceleration —
• Fusion Complexity Is Not Fusion Quality: Backbone, Cross View Fusion and Metadata as Separate Factors in Pasture Biomass Regression —
• LMEB: Long-horizon Memory Embedding Benchmark —
• DirPA: Addressing Prior Shift in Imbalanced Few-shot Crop-type Classification —
• QTrack: Query-Driven Reasoning for Multi-modal MOT —
• Deep probabilistic model synthesis enables unified modeling of whole-brain neural activity across individual subjects —
• Predicting Stress-strain Behaviors of Additively Manufactured Materials via Loss-based and Activation-based Physics-informed Machine Learning —
• AURORA-KITTI: Any-Weather Depth Completion and Denoising —
• HYDRA: Representation Harmonized Tokenization for Multimodal Generation and Understanding —
• TokenPB: Enabling VLA Action Self-Verification via VLM Token Probability Bucketing —
• Probing and Inducing Clinical Modality Grounding in Brain Tumor Segmentation —
• Learning Dynamic Belief Graphs for Theory-of-mind Reasoning —
• The Hidden Puppet Master: Predicting Human Belief Change in Manipulative LLM Dialogues —
• Multifidelity-Augmented Gaussian Process Inputs for Surrogate Modeling from Scarce Data —
• GIFT: Global Irreplaceability Frame Targeting for Efficient Video Understanding —
• JAL-Turn: Joint Acoustic-Linguistic Modeling for Real-Time and Robust Turn-Taking Detection in Full-Duplex Spoken Dialogue Systems —
• Weakly Convex Ridge Regularization for 3D Non-Cartesian MRI Reconstruction —
• Logarithmic Scores, Power-Law Discoveries: Disentangling Measurement from Coverage in Agent-Based Evaluation —
• Soft MPCritic: Value-Based Reinforcement Learning for Non-Differentiable Predictive Control —
• Satellite-Free Training for Drone-View Geo-Localization —
• Exploring Motion-Language Alignment for Text-driven Motion Generation —
• When Rewards Disagree: Reward Weighting for Token-Efficient Video Reasoning —
• Mitigating Distribution Sharpening in Math RLVR via Distribution-Aligned Hint Synthesis and Backward Hint Annealing —
• YIELD: A Large-Scale Dataset and Evaluation Framework for Information Elicitation Agents —
• MosaicMRI: A Diverse Dataset and Benchmark for Raw Musculoskeletal MRI —
• PR-MaGIC: Prompt Refinement Via Mask Decoder Gradient Flow For In-Context Segmentation —
• ReasonXL: A Large-Scale Corpus for Language-Specific Reasoning in LLMs —
• KMMMU: Evaluation of Massive Multi-discipline Multimodal Understanding in Korean Language and Context —
• Feed-Forward 3D Scene Modeling: A Problem-Driven Perspective —
• Faithfulness Serum: Mitigating the Faithfulness Gap in Textual Explanations of LLM Decisions via Attribution Guidance —
• Physically-Induced Atmospheric Adversarial Perturbations: Enhancing Transferability and Robustness in Remote Sensing Image Classification —
• The Courtroom Trial of Pixels: Robust Image Manipulation Localization via Adversarial Evidence and Reinforcement Learning Judgment —
• EmbodiedHead: Real-Time Listening and Speaking Avatar for Conversational Agents —
• MedPRMBench: A Fine-grained Benchmark for Process Reward Models in Medical Reasoning —
• Mira-Embeddings-V1: Domain-Adapted Semantic Reranking for Recruitment via LLM-Synthesized Data —
• Local-First Intelligence via Proactive SLM Collaborating with Feedback-Providing LLM —
• Negative Advantages Is a Double-Edged Sword: Calibrating advantages in GRPO for Search Agents —
• Optimizing Data Augmentation for Real-Time Small UAV Detection: A Lightweight Context-Aware Approach —
• Cover meets Robbins while Betting on Bounded Data: n Regret and Almost Sure n Regret —
• PermaFrost-Attack: Stealth Pretraining Seeding(SPS) for planting Logic Landmines During LLM Training —
• Cross-Stage Coherence in Hierarchical Driving VQA: A Consistency Protocol and Parameter-Efficient Gated Context Routing —
• From graphemic dependence to lexical structure: a Markovian perspective on Dante's Commedia —
• The Dynamics of Delusion: Modeling Bidirectional False Belief Amplification in Human-Chatbot Dialogue —
• 3D-LENS: A 3D Lifting-based Elevated Novel-view Synthesis method for Single-View Aerial-Ground Re-Identification —
• ControBench: An Interaction-Aware Benchmark for Controversial Discourse Analysis on Social Networks —
• Structure Liberates: How Constrained Sensemaking Produces More Novel Research Output —
• Dependency Parsing Across the Resource Spectrum: Evaluating Architectures on High and Low-Resource Languages —
• Are LLMs Ready for Conflict Monitoring? Empirical Evidence from West Africa —
• The Shared Latent Structure of LLM Psychometric Responses —
• AeroJEPA: Learning Semantic Latent Representations for Scalable 3D Aerodynamic Field Modeling —
• XL-SafetyBench: A Country-Grounded Cross-Cultural Benchmark for LLM Safety and Cultural Sensitivity —
• Training-Free Dense Hand Contact Estimation with Multi-Modal Large Language Models —
• When Does Trimming Help Conformal Prediction? A Retained-Law Diagnostic under Calibration Contamination —
• SphereVAD: Training-Free Video Anomaly Detection via Geodesic Inference on the Unit Hypersphere —
• HIMMEL: Hierarchical Interleaved Multi-stream Motion Encoding for Long Video Understanding —
• Source or It Didn't Happen: A Multi-Agent Framework for Citation Hallucination Detection —
• RewardHarness: Self-Evolving Agentic Post-Training —
• ErrOPD: When Does On-Policy Distillation Transfer Teacher Errors? —
• Soohak: A Mathematician-Curated Benchmark for Evaluating Research-level Math Capabilities of LLMs —
• PGID: Progressive Guided Inversion and Denoising for Robust Watermark Detection —
• Overcoming Catastrophic Forgetting in Visual Continual Learning with Reinforcement Fine-Tuning —
• Federated Language Models Under Bandwidth Budgets: Distillation Rates and Conformal Coverage —
• Dynamic Skill Lifecycle Management for Agentic Reinforcement Learning —
• Revealing the Gap in Human and VLM Scene Description through Counterfactual Semantic Saliency —
• Concurrency without Model Changes: Future-based Asynchronous Function Calling for LLMs —
• alpha-TCAV: A Unified Framework for Testing with Concept Activation Vectors —
• Conditioned in Training, Dispensable at Inference: Typological Conditioning for Multilingual Speech Translation —
• PQR: A Framework to Generate Diverse and Realistic User Queries that Elicit QA Agent Failures —
• FrequencyBooster: Full-Frequency Modeling for High-Fidelity Pixel Diffusion —
• Training data attribution in diffusion models via mirrored unlearning and noise-consistent skew —
• deadtrees.earth-aerial: A Multi-Resolution Aerial Image Dataset for Tree Cover and Mortality Detection —
• OScaR: The Occam's Razor for Extreme KV Cache Quantization in LLMs and Beyond —
• SurgOnAir: Hierarchy-Aware Real-Time Surgical Video Commentary —
• Self-Policy Distillation via Capability-Selective Subspace Projection —
• VideoOdyssey: A Benchmark for Ultra-Long-Context and Omni-Modal Video Understanding —
• CRONOS: Benchmarking Counterfactual Physical Consistency in Video Models —
• Plume Segmentation from MethaneSAT with Cross-Sensor Transfer Learning and Physics-Informed Postprocessing —
• Does Seeing More Mean Knowing More? Mono-Anchored Advantage Normalization for Multi-Source Visual Reasoning —
• Selective Latent Thinking: Adaptive Compression of LLM Reasoning Chains —
• Pixel-Level Pavement Distress Assessment Using Instance Segmentation —
• Beyond Questions: Evaluating the Stated Knowledge of LLMs —
• How Do Uncertainty Estimators Align with Hallucination Occurrence in LLMs? —
• Escape the Language Prior: Mitigating Late-Stage Modality Collapse in Audio Reasoning via Modality-Aware Policy Optimization —
• Self-Improving Language Models with Bidirectional Evolutionary Search —
• Trajectory Constraints for Imaging Inverse Problems —
• Learning Representations from 3D Gaussian Splats —
• Masked Diffusion Vision-Language Models for Temporal Action Localization —
• Conformal Certification of Reasoning Trace Prefixes —
• Cross-Lingual Steering for Figurative Language Generation —
• Hoeffding Concept Bottleneck Models with Applications to Overhead Images —
• Real2SAM2Real: Generative 3D Caches as Complementary Context for Video Diffusion —
• Learning to Retrieve: Dual-Level Long-Term Memory for Text-to-SQL Agents —
• Agentic Clustering: Controllable Text Taxonomies via Multi-Agent Refinement —
• Better with Experience: Self-Evolving LLM Agents for Evidence-Grounded Health Community Notes —
• Learning When to Translate for Multilingual Reasoning —
• SkillHarm: Lifecycle-Aware Skill-Based Attacks via Automated Construction —
• Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactions —
• MLP Splatting: Object-Centric Neural Fields —
• NoRA: Evaluating Grounded Reasonableness in Visual First-person Normative Action Reasoning —
• Plan, Watch, Recover: A Benchmark and Architectures for Proactive Procedural Assistance —
• Improving Heart-Focused Medical Question Answering in LLMs via Variance-Aware Rubric Rewards with GRPO —
• Ultra Flash: Scaling Real-Time Streaming Video Generation to High Resolutions —
• VISTA: A Versatile Interactive User Simulation Toolkit for Agent Evaluation —
• P3D-Bench: Benchmarking MLLMs for Parametric 3D Generation and Structural Reasoning —
• DarkVGGT: Seeing Through Darkness Using Thermal Geometry —
• Task-Aligned Stability Analysis of Vision-Language Models for Autonomous Driving Hazard Detection —
• FORT-Searcher: Synthesizing Shortcut-Resistant Search Tasks for Training Deep Search Agents —
• Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning —
• Understanding Helpfulness and Harmless Tension in Reward Models —
• What Drives Test-Time Adaptation for CLIP? A Controlled Empirical Study from an Update Perspective —
• Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning —
• Conditional Multi-Event Temporal Grounding in Long-Form Video —
• A Decision-Theoretic View of Test-Time Training: When, How Far, and Which Directions to Adapt —
• LLM Judges Have Dark Current: Stratified Null and Utility Tests for LLM Judge Reliability —
• Retrievable Gradients: Continual Post-Training Without Cumulative Weight Drift —
• Learning a Sampling-Free Variational DNN Plugin from Tiny Training Sets to Refine OOD Segmentation With Uncertainty Estimation —
• Text-Vision Co-Instructed Image Editing —
• PseudoBench: Measuring How Agentic Auto-Research Fuels Pseudoscience —
• Beyond Reward Engineering: A Data Recipe for Long-Context Reinforcement Learning —
• Decoupling Search from Reasoning: A Vendor-Agnostic Grounding Architecture for LLM Agents —
• STAGE: Source-grounded Text-to-JSON Artifact GEneration —
• Through the PRISM: Preference Representation in Intermediate States of Video Diffusion Models —
• Horizon Adaptive Offline Policy Learning via Value Stitching —
• LLM and Human Modes of Representation —
• Plans Fade Fast: Why Context Management Is Load-Bearing for LLM Agents —
• Safe Few-Step Generation via Velocity Editing —
• Assessing Post-Reform Changes in Risk Disclosure Quality with a Multidimensional Text Analysis Approach —
• Mapping Political-Elite Networks in Europe with a Multilingual Joint Entity-Relation Extraction Pipeline —
• Error-Conditioned Neural Solvers —
• SPARK: Adaptive Parametric Knowledge Participation for Faithful Retrieval-Augmented Generation —
• MedEvoEval: Evaluating Continual Evolution of Doctor Agents through Simulated Clinical Episodes —
• Synthetic CT as a First Safeguard for Choosing Medical VLMs in Lung Cancer Screening —
• CylindTrack: Depth-Aware Cylindrical Motion Modeling for Panoramic Multi-Object Tracking —
• Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs —
• GMO-E squared DIT: Grounded Multi-Operation Editing for E-Commerce Images —
• Beyond Heatmaps: Unsupervised Concept-Graph Reasoning for Interpretable Visual Explanation —
• Transformers with Physics-Informed Encodings and Simulation-Based Inference for Robust Detection of Eccentric Binary Black Holes in Pulsar Timing Array Data —
• Qantara: Bridge-Flow Training for Multi-Paradigm JEPA Inference —
• CenSynCMB: Centre Maps and Physics-Guided Synthesis for Microbleed Detection —
• LLMs Silently Correct African American English: Auditing and Mitigating Dialect Bias via Activation Steering —
• Automatic Thematic Indexing of Large Literary Corpora: A Machine Learning Approach to Voltaire's Complete Works —
• DGSfM: Depth-Guided Scale-Aware Global Structure-from-Motion —
• Scaling Point-in-Time Language Models: Enabling Economic Evaluation of Embeddings —
• Tracing Agentic Failure from the Flow of Success —
• From hyperplanes to hyperellipsoids: a pedagogical bridge from linear classifiers to single-qubit mixed-state classifiers —
• Planner-Guided Recursive Exploration for Diffusion Language Model Decoding —
• IoU-PD: IoU-Aware Privileged Distillation for Visual Grounding with Multimodal Large Language Models —
• DRNOISE: Benchmarking Deep Research Agents in Misleading Evidence Environments —
• H squared SD: Hybrid Hindsight Self-Distillation —
• Task Competence Is Not Instruction Following: Evaluating Instruction-Conflicting Behavior in Small Language Models —
• Exposure is Optional: Learning Unlike Coordination in Language Models —
• Activation Steering with Jungian Cognitive Functions —
• This Is a Good Paper, Right? How Tag Questions Affect Sycophancy Across 106 Language Models —
• ForgetBench: Benchmarking Forgetting Dynamics of Long-Term Parametric Memory in Language Models —
• AtmosERC: Modeling Dialogue-Level Affective Atmosphere for Emotion Recognition in Conversation —
• Same Illness, Different Voice: Restating the Patient's Account Narrows Narrative Anchoring in Clinical Language Models —
• LEMUR: Learning to Align with Multi-Objective Reinforcement Learning from Preference Feedback —
• CORTIVA: Candidate-Score Fusion of Complementary Visual Teachers for EEG- and MEG-to-Image Retrieval —
• Qwen-3D: A Generalist 3D Vision-Language Model for Spatial Understanding —
• Interoceptive Attention as Dynamic Homeostatic Prioritization in a Foraging Agent —
• Marrying Optimal Transport and ODEs for Unified Continuous-Time 4D Reconstruction and Tracking —
• MAD-HOI: Masked Autoregressive Diffusion for Generating Articulated Hand Object Interactions from Text —
• Multi-View Relational Distillation for Spatial Reasoning with Vision-Language Models —
• GALA: Generation-Aware Cross-Modal Alignment for Text-to-Time-Series Synthesis —
• Scaling Creative Writing Beyond Story-Centric Data with Attribute-Guided Genre Expansion —
• Time as Structure: Temporal Dependency Graphs for Verifiable Deadline Computation over Legal Documents —
• Certified but Private: Scalable Zero-Knowledge Proofs for Neural Network Guarantees —
• Write, Execute, Refine: From Skill Follower to Skill Optimizer via Reinforcement Learning from Execution Feedback —
• SIGMA: Symmetry-aware, Intelligent, Geometric, Multi-faced Adaptive Traffic Control for Reliable Intelligent Transportation Systems —
• Identifying Implicit Premises for Logical Reconstruction of Argument Graphs —
• Remember, Verify, or Ask? Evaluating Memory Decisions at Reuse Time —
• SafeBranch: Branch-Pair Safety Alignment for Embodied Agents —
• ParallelWorld: Test-Time Scaling for Agentic Embodied Reasoning —
• AnaDiffusion: Anatomically Compositional Latent Diffusion for Controllable 3D Brain MRI Generation —
• OptiGeo: Efficient Monocular Geometry for Embodied Perception in Optically Challenging Scenes —
• RingMoClaw: An Experience-Inspired Multi-Agent Framework for Self-Evolving Research in Remote Sensing —
• Evidential Deep Learning for Multi-Modal Anti-UAV Detection —
• SliceBridge: Context-Consistent Repair of Corrupted Slice Intervals in T1-Weighted MRI —
• TAPVid-MV: A Benchmark for Tracking Any Point in 3D Across Multiple Views —
• Balancing Frequencies and Pixels in Flow Matching —
• Selective Knowledge Control for Continual GUI Agent Learning over Application Streams —
• Programmable World Model —
• SCINTILLA-SNN: A Spiking Multi-Scale Selective Aggregation Network for Perineural Invasion Prediction —
• V-ICAL Bench: Evaluating Video In-Context Learning for Multimodal Agents in Interactive Environments —
• MEgoVista: Multi-view Ego-aware Motion Estimation for Metric 4D Hands and Head in the Wild —
• QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents —
• TRACE: Expert-Aligned ECG Representation Learning with Rigorous Benchmarking and Real-World Validation in Acute Cardiac Care —
• A Differentiable Optimization Framework for Registering Sequential Bounding Boxes with Point Cloud Stream —
• SlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving —
• MRFlow: Mechanism-Structured Residual Flow Matching for Virtual Cell Perturbation Prediction —
• PRIME: Frequency-Aware Unified Pruning and Merging for MoE Compression —
• PsyBer-Agent: A Psychology-Driven User Agent for Recommender Systems —
• How Vision Transformers Represent Position: A Mechanistic Comparison of Absolute and Rotary Positional Encodings —
• Posterior Flow Matching —
• GameSmith: Scaling Zero-Sum Self-Play with Synthesised Games —
• SharedSAE: a unified sparse dictionary across u-net blocks and denoising time for diffusion models —
• SemTree: Judge-Aware Semantic Decision Trees for Inspectable and Repairable Rule Decisions —
• Artificiality's First Exam: An LLM-Referenced C-test for Supraintelligence —
• Barycentric Anchored Weight Search for Robust Model Merging —
• Contrastive Covariance Subspaces for Activation Steering —
• Can We Trust This Evaluation? Instance-Level Reliability of LLM-as-a-Judge —
• Rotate the Loop: Blockwise Conformality for Looped Transformers —
• PLDiv: Measuring Dataset Diversity and Semantic Collapse via Persistence Landscapes —
• A Geometric Perspective on Reward Function Updates in Inverse Reinforcement Learning —
• Loose Lips: Understanding Privacy Leakage in Agentic Dialogue —
• Is the Future Predictable? Modeling Future Predictability for Reliable Temporal Reasoning —
• Pristine Models —
• Human Pose Estimation by Probabilistic Mixtures of Gaussian and Uniform Distributions —
• Scenes as an Attack Surface: Uncovering Scene-Grounded Safety Failures in MLLMs —
• Fin Agent at FinEval 2026: Task-Aware Methods for Multilingual Financial NLP —
• Does System One Transfer? Calibrated Typed-Decision Models, Label Budgets, and Translation for Indonesian Regional Languages —
• ClawLoop: Self-Improving Agents via Harness-Native Training Loop —
• A Stochastic Analysis Perspective on Diffusion Models —
• TrajCFG: Token-Wise Classifier-Free Guidance via Denoising Trajectory Feedback —
• Video-Arena: A Scalable Agentic Video Benchmark for Spatiotemporal Capability —
• Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation —
• Unconstrained CTI-REALM with Open-Weights —
• Token-Budgeted Candidate Retrieval for 3D Scene-Grounded VLMs: Category Names vs. Pairwise Geometry —
• Agreement versus Softmax under Spatial Disambiguation —
• Modeling the Arrival Time of Information-Bearing Events in Natural Language —
• Learnable Low-Rank Polynomial Sketch for Effective Linear Attention —
• Harness 2: Recursive Agent Harnessing for an Open World —
• CAPTOR: a vision-language foundation model for Chest-Abdomen-Pelvis Oncologic CT —
• Latent World Action Model for GUI Agents —
• Agentic Data Engineering for LLMs: System Design and a Controlled Empirical Study —
• Learning Non-Gaussian SEMs with Conditional Risk, and Why the Direction of Recursion Matters —
• SliderEdit++: Improved Slider Trajectories in Instruction-Based Image Editing —
• Image Diffusion Models Are Good 3D Human Motion Generators —
• Variance Fingerprints: Component-Wise Identification of Content Factors by Augmentation Design —
• BigCell: Generating Gigapixel Whole-Slide Images —
• Thinking with Images without Coordinates —
• RoboFake: Are Deepfake Detectors Ready for Humanoid Face Robots? —
• The Thinking Leakage in Reinforcement Learning for LLMs —
• TGSR: Text-Grounded Diffusion for Faithful Image Super-Resolution —
• GeoFix: Geometry Fixing for 4D Consistent Generation —
• Controllable Diverse Sampling through Pullback Directions —
• Learning to See through Action: Predictive Curiosity for Robots —
• ConicFlow: constrained Inference-time optimal control for flow matching models —
• ASSET: Adaptive Speech Segmentation via Reinforcement Learning —
• MvFFN: Multi-view Floor plan Feed-forward Network for Unposed Wide-Baseline Panorama Layout Reconstruction —
• EV-JEPA: Elastic V-JEPA for Any Device and Cross-Device Inference —
• Manifold Navigation Learning Using Recurrent Networks: Efficient Training Techniques —
• Real-Time Traffic Density Estimation: A Hybrid Computer Vision and Predictive Machine Learning Pipeline —
• TraceSampler: Making Sampling Errors Predictable and Correctable for Accelerated Generative Sampling —
• Which Metadata Matters for LLM Tool Selection in MCP? Evidence from Controlled Factor-Level Interventions —
• Continuous-Time Cloth Dynamics with Physics-Informed Neural ODEs —
• Evaluating LLMs as Agentic Workflow Executors in Real-World Customer Support Skills —
• Multimodal Co-Generation With Continuous Bitstream Diffusion —
• Learning Transferable World-Action Models from Task-Paired Human Videos —
• History, Adaptation, and Hidden Gain: Two Prospective Tests of Learned Predictive Control in a Synthetic Neural Population —
• Decomposing A Neural Learner —
• SciCurateBench: Curating Experimental Data from Heterogeneous Scientific Literature with Verbal-Action Agents —
• RealityBench: Simulated Environments to Evaluate a Vision Language Model's Ability to Document Human Physical Work —
• VisualCoding: Learning Executable Skills for Reliable Visual Reasoning —
• Scaling Laws for Multimodal Data Mixtures —
• Design Choices for Multimodal Transformer Learning with Missing and Longitudinal Healthcare Data —
• Neither True nor False: Do LLMs Represent Truth-Value Gaps in Plural Predication? —
• SceneForesee: Learning Scene Evolution for Spatiotemporal Reasoning —
• Evaluating Semantic Relation Knowledge through Bidirectional Relation Prediction from Gloss Pairs —
• FreeCHROMA: Generative Chroma Keying from Unaligned Background References —
• Evaluating Bias in Chinese Large Language Models across Tasks and Registers: Task Operationalization Effects and Irony Classification Risks —
• EdgeLLM-Nano: Auditing Polar-Guided Key-Cache Compression and Consistency-Anchored Pruning on a 33.6 MB On-Device Language Model —
• Scale, Drivers, and Limits of Representational Alignment Across Large Language Models —
• AphaVoice: Text-to-Speech Model for Aphasia Patient Simulation —
• DSR-Net: Learning Dynamic Skeletal Relations for 3D Human Pose Estimation —
• Spatiotemporal Recognition of Abdominal Wall Layers During Optical Trocar Entry —
• TartanMatch: Towards Universal Dense Correspondence Across Modalities —
• Compositional Concepts Erasing: Disentangling Object-Style via Domain Manifold Unlearning —
• We Have Been Interpreting Diffusion Transformer Backbone Behaviors Wrong —
• Query-Adaptive Neuron Steering for LLMs —
• Identication and Estimation of Nonsymmetric Determinantal Point Processes —
• ScienceBuddy-PV: A Verifier-Governed Recursive Improvement Protocol for Panvascular Intervention Research —
• LongHisDoc: A Comprehensive Benchmark for Chinese Long Historical Document Understanding —
• InfraWorld: Motion-Aware World Modeling for Roadside 4D Occupancy Forecasting —
• Adversarial Gradient Forcing for Robust Few-Step Autoregressive Video Generation —
• Strong Gaussian Approximation of Constant-Learning-Rate SGD: Bias and Trajectory Inference —
• FrameFold: Response-Guided KV Cache Consolidation for Autoregressive Long Video Generation —
• Reward-Aligned Distillation —
• Uniform Information Density-Based Preference Optimization —
• Attribute-Centric Vision-Language Alignment for Fine-Grained Gait Understanding —
• Bidirectional Adaptive Transformers for Multimodal Anomaly Detection —
• Decoupling Knowledge Forgetting and Refusal Learning in MLLM Unlearning —
• EviScene: Beyond "Is It AI?" for Agentic Image Factuality Assessment —
• EarthAlign: Learning Language-Aligned Representations of Earth System Data —
• DuSRFlow: Dual-Lens Super-Resolution via Parallax-Aware Flow Estimation and Complementary Reference Warping —
• Automatic assembly modeling through attention-based assembly feature mate prediction and assembly feature pattern recognition —
• PVFH-Bench: Benchmarking Premise-Induced Visual Fixation Hallucinations in Vision-Language Models —
• Can Synthetic UAV Benchmarks Be Trusted? Real-Synthetic Validation for UAV Model Selection —
• TAG: Calibrating SAM3 for Universal Image Segmentation without Dense Supervision —
• Prospective Forcing: Learning to Look Ahead for Autoregressive Long Video Diffusion —
• TGuide: Faster Autoregressive Video Diffusion via Historical Trajectory Guidance —
• TETRIS-4DGS: Efficient Post-Training Compression of 4D Gaussian Splatting —
• SPARE: Shared Perception and Adaptive Retrieval of Evidence for Streaming Video Understanding —
• Competition Domains Matter: Selective Head Globalization for Training-Free Sparse Attention in Video Diffusion Transformers —
• CAST-ASD: Causal Attention over Speakers and Time for Online Active Speaker Detection —
• HumanPhys: A Human-Centric Physical-World Perception Benchmark for Real-World Visual Reasoning —
• A Hybrid Kinematic GCN and Elastic Transformer for Efficient Indian Sign Language Recognition —
• Demiurge: Forging Navigable 4D Worlds via Autoregressive Video Rerendering —
• Understanding Hidden Spatial Preferences in Bayesian Optimization —
• PURE: Refining Feed-forward 3D Reconstruction with Unlabeled Post-training —
• Lingdan-V2: Large language models for clinically grounded reasoning in traditional Chinese medicine —
• How to Build a Benchmark That Actually Measures Taste —
• Vela: User Behavior as Supervision Sparks LLM Intelligence —
• PixARG: Pixel-Asynchronous Retrieval-Augmented Diffusion for Fast and Realistic Image Generation —
• LightVAE: Towards Compact and Efficient Video Autoencoders —
• Align-VAE: Aligning Attention and Geometry in VecSet VAEs for 3D Reconstruction —
• Japanese Handwriting Generation Based on Human Preference Feedback Using Direct Preference Optimization —
• Amharic-English Cross-Language Question Answering Using Statistical Machine Translation, Named Entity Recognition, and Semantic Indexing —
• The Constitution Speaks Nheengatu: An Open MT System and Reproducible Corpus Pipeline for the Amazonian L'ingua Geral —
• CUSE: Fine-Grained Multimodal Unlearning with Counterfactual Guidance and On-Policy Self-Distillation —
• Who Receives Respect? A Counterfactual Benchmark for Honorific Allocation Bias in Bengali —
• Training-Free BioNEL-RAG: Advancing Biomedical Entity Linking through Retrieval-Augmented Generation and Large Language Model —
• YOUDrive: Driving Style as a Composable Direction for End-to-End Autonomous Driving —
• [Re] Boosting the Visual Interpretability of CLIP via Adversarial Fine-Tuning —
• Physics-informed random neural architectures on compact boundaryless multiply connected manifolds for random-operator models in machine learning and uncertainty quantification —
• Towards Privacy-Preserving Thermal Human Perception from Dataset to Deployment —
• Getting Space Right: Agentic Rewards for Image Generation —
• CAST: Real-Time Motion Capture for Any Skeleton Topology —
• ScalLoc: Continuous Oblique Thermal UAV Geo-Localization with a 2D Satellite Map —
• MORE: Learning Motion-Oriented Registers for Driving World Models and Planning —
• Dependency Reasoning for Aspect Sentiment Quadruple Prediction —
• Social Bias in Chinese Large Language Models under Social Media Expressions —
• Agent-Guided Evidence Acquisition for Emotion-Cause Pair Extraction —
• TeachGrade: Training-Free Short Answer Scoring with Teacher Evidence —
• Prediction with Expert Advice under Heavy-Tailed Losses —
• Minimum Volume Optimal Transport for Multivariate Conformal Prediction —
• WorkBorne: Discovering Agentic Tasks from Real-World Artifact Transitions —
• SkillForgeBench: Evaluating Agent Reliability Grounded in Agent Skill Specifications —
• FidelityQ: Preserving Predictive Fidelity in Ultra-Low-Bit Language Model Quantization —
• Predicting how model propensities generalize out of distribution —
• Engram Neuron Inspired Adaptive Memorization for Lifelong Learning —
• Compositional Average Velocity Modeling for Efficient Diffusion Offline Reinforcement Learning —
• Text-Guided Medical Image Segmentation via Residual Aggregation and Dense Alignment Representations —
• DSM-Free UAV Visual Localization via Latent-Depth Variable Projection and Structure-Adaptive Coupling —
• Do Personalized LLMs Need the Whole Conversation? Matching the Full History with Only 5% of Its Text via Relevance-Filtered Context —
• CatPortrait: Taming Video Generative Models for Portrait Animation via In-context Conditioning —
• What Survives When Activation Steering Is Used as Memory —
• AlphaRLM: Aligning Language Models for Literature-Grounded Symbolic Factor Discovery —
• HiPAD: A Unified Gait, Audio-Visual, and Personality Fusion Network for Depression Detection, with a Hierarchical Personality Ablation and the Calibration Limits of Domain-Adapted LLMs —
• AD-GS: Anatomical Density-guided 3D Gaussian Splatting for Sparse-View Dental CBCT Reconstruction —
• HodgePhaseNet: Discrete Hodge Decomposition of Semantic Flows in Transformer Attention —
• Metric-Driven Reinforcement Learning for Object Hallucination Reduction in LVLM-Based Image Captioning —
• Mixing Regularization for Average-Reward Q-Learning and Q-Planning —
• Characterizing the Separability Gap in Prior-Data Fitted Networks —
• Retraining-Free Depth Pruning of Vision Transformers: A Cross Domain Study of Layer Redundancy —
• CowCare: A Unified Visual Framework for Viewpoint-Robust Cattle Tracking and Spatiotemporal Heat-Stress Analysis —
• UltraBERT: Beyond the Limit of BERT Pre-Training —
• Toward System One Clinical Intelligence: Parameter-Efficient Fine-Tuning and Counterfactual Explainability of Decision Foundation Models —
• Appearance Compensation in Volume Rendering —
• What Really Works for Universal Multimodal Embeddings? A Controlled Study of Contrastive Training Strategies —
• Sparsity Saves Compute, Not Samples: Oracle Inequalities for Dense and Sparse Mixture of Experts —
• PAT: A Personality-Driven Augmentation and Transfer Learning Framework for Depression Detection on Social Media —
• Use Of Convolutional Neural Networks In Detection Of Branch Dieback Disease In Pequi Trees —
• Select or Rewrite: Oracle-Guided Fine-Tuning for Generative ASR Post-Editing from N-Best Hypotheses —
• Deep Learning-based Semantic Segmentation of Dental Radiographs: Comparative Evaluation of PSPNet, DeepLabv3, UNet, and UNet++ —
• Global Disaster Risk Assessment Using World Risk Index Data: A Geospatial and Visual Analytics Approach —
• BrainAid: Efficient Brain Tumor Detection Using Lightweight CNNs and Explainable AI for Resource-Constrained Environments —
• Region Aware Squeeze-and-Excitation Networks for Image Segmentation —
• Have You Been Convinced? A Neuro-symbolic Argumentation Reasoning Approach to Detecting Persuasiveness in ChangeMyView Discussions —
• VLCQ: Post-training quantization for deep neural networks using variable length coding —
• TIDE: Trajectory-Informed Skill Distillation for Efficient Text-to-Video Prompt Optimization —
• RegExtract: A Benchmark for PFAS Regulatory Information Extraction with LLMs —
• Drive3R: Block-wise Streaming Geometric Foundation Models for Autonomous Driving —
• Data-Topology Compatibility Enables Personalization in Decentralized Learning —
• RIECNN: Real-Time Image Enhanced CNN for Traffic Sign Recognition —
• Lumen-bench: Cross-Lingual Safety Gaps and the Say-Do Problem in Tool-Calling Large Language Models —
• Compute-Optimal Scaling Laws for Spiking Language Models —
• Can agents whose sole duty is to keep humans alive be hijacked to kill? Delayed multi-turn derived-action trap attacks in simulated high-stakes environments —
• Where Did We Go Wrong? Localizing Errors to Repair Multi-Turn Conversations —
• Learning Proposition-Level Links from Sentence-Level Supervision for Neuro-Symbolic NLI —
• Sanzaru: Benchmarking Protocol Formation under Epistemic & Communication Asymmetry —
• TRIG: Toward Measuring Texture Realism in Image Generation —
• Active Learning for Finding a Needle in a Haystack —
• ExeWorldBench: Benchmarking Code-Based Executable 3D World Modeling —
• Only What You've Signed up For: Secure Workflows with Constrained Diffusion Agents —
• UDHRBench: Benchmarking Human Rights in Large Language Models —
• Game-Theoretic Reinforcement Learning based Topology-Agnostic Defense for Virtual Power Plants —
• SAFE-KV: Causal Suffix Repair for Behavior-Preserving Non-Prefix KV Cache Reuse in RAG —
• Do LLMs differentiate linguistic traits internally?: A layer-wise analysis —
• Exploring Jailbreak Vulnerability Across Depth-Based Dynamic Inference in Large Language Models —
• PANINI-DR: Controlled Deep Research through Generative Semantic Workspaces —
• Sequent-Prover: Training Agents for Formal, Checkable SMT-based Reasoning —
• LeHami: Symplectic Latent Dynamics for World Models —
• Do correct answers have stable latent knowledge support? —
• GeoBridge: Evolving Transferable Decoder Topologies for Geometric Vision —
• Imperfect Verifiers, Exact Targets: Inference-Time Scaling for Continuous Generation —
• North or South? Diagnosing LLMs' Understanding of North Korean Language, Society, and Knowledge —
• What a Data Release Must Preserve for Off-Policy Evaluation —
• Protein Crystallization Segmentation and Classification Using Subordinate Color Channel in Fluorescence Microscopy Images —
• ASM-6D: Fast Certified Pose-Shape Inference with Complete Correspondence Flow Maps —
• HGA-Quant: Heterogeneity-Aware Quantization for AutoRegressive Visual Generation Models —
• Accuracy and Uncertainty in Machine Deception Detection (Research White Paper) —
• Channel Match: Calibrating Shape Latents for Few-Step 3D Generation —
• TempoKV: Predictive KV Cache Quantization for Autoregressive Video Generation —
• CIViC: Evidence Gates for Audio-Conditioned Visual Prediction on Proxy Tasks —
• BaT-SAM: Introducing Inductive Bias into SAM via a Mixture of Convolutional Experts for Brain Tumor Segmentation —
• StimFlow: Synthesizing Brain Activity from Images via Stimulus-conditioned Flow Matching —
• Across Viewpoints and Time: Trajectory-Generalizable Physical Attacks on World Models and Video Understanding Models —
• Video-Based Sexual Harassment Detection Using Fine-Tuned Video Transformers —
• UniForge3D: Joint 3D Reconstruction and Generation in a Shared Anchor-Guided Space —
• A Sovereign, Indic-First, Reasoning Model —
• In-Place Indic Tokenizer Expansion and Continual Pretraining of Nemotron-3-Nano-30B-A3B —
• VideoEvoHarness: Self Evolving Agent Harness from Unlabeled Videos for Video Understanding —
• Harnessing VLM in Embodied RL: From Reward Regressor to Success Detector via Exploration-Augmented Key-Observation Gating —
• A Decade of PhotoMappers: A Longitudinal Study of Crowdsourced Disaster Photos Geolocalization with Vision-Language Models and Geospatial Reasoning —
• From Candidate Tokens to Final Outcomes: Decoding Relative Long-Horizon Candidate Value from Reasoning Model Representations —
• OmniRSI: Taming Latent Diffusion for Unified Multimodal Remote Sensing Image Generation —
• Sensitivity or Criterion? Measuring and Moving When Language Models Act —
• Harnessing SPAD Noise: Test-Time Gated Attention for Indoor Depth Estimation —
• Rethinking Supervision in Document Unwarping: A Self-consistent Flow-free Approach —
• Hidden in Plain Sight: Advancing Harmful Content Understanding via Perception-Aware Adversarial Self-Play —
• EmoTrace-Bench: Diagnosing Emotional Expression in Text-to-Video Generation through Controlled Cues and Intensity —
• Semantic Saliency from Multi-Modal Large Language Model Scene Understanding Maps —
• AlphaPilot: Multi-Turn Tool-Using LLM Agents for Adaptive Alpha Mining with Reinforcement Learning —
• Concept Representations in Vector Space: Quantifying Sharedness and Idiosyncrasy of Meaning —
• Structured Stateful Relational Scripts with a High-Quality Dataset for Multi-Shot Audio-Visual Video Generation —
• REPRO: Reflective Evolutionary Prompt Optimizer for Automated LLM Red-Teaming —
• SARATHI: Decoupled Data-Free Probing and Adaptive Reconstruction for Structured LLM Pruning —
• Semantically Steerable Image Tokenization —
• A Step in the Right Place: Understanding and Optimizing Timestep Placement in Continuous Diffusion Language Models —
• UMAS3D: Unified Mask-Free 3D Editing with Region-Adaptive Source Preservation —
• Apertus 1.5: From Open Text Models to Open Multimodal Reasoners —
• An Interpretable Linguistically-Grounded Analysis of Scene Saliency in Movie Screenplays —
• ELECT: Settling Identity Before Rendering for Scalable Story Visualization —
• A Plausible Future Is Not a Causal Future: Can Video Models Generate How the World Unfolds? —
• Genesis: Visual Augmentation Primitives for Multimodal Chain-of-Thought Reasoning —
• Truncation Samplers Are Nested: Comparing Decoding Rules Is a Measurement Problem —
• Latent-Space Importance Weighting under Covariate Shift —
• The Fragility and Dynamics of LLM Pruning-Recovery Steps —
• Towards Robust Calibration Dataset for Domain-General Mixture of Experts Pruning —
• Ellipsoids from Tangents —
• The Gradient That Top-k Distillation Never Computes —
• FireNav: Decision-Oriented Reconstruction Evidence Routing for Vision-Language Navigation in Indoor Fire Environments —
• Squeeze-Aware-YOLO: Early Channel Recalibration and Multi-Scale Fusion for Real-Time Traffic Sign Detection —
• Lesion-Aware Attention for Mammography Segmentation with Heterogeneous Supervision —
• Joint Contrastive-Adversarial Representation Learning for Unsupervised Domain Adaptation —
• Screening Collapse under Recursive Synthetic Data —
• 1Step-WAM: Your World Action Model but Faster using 1-Step Action Generation —
• Robust Detection of Directional Adversarial Attacks in Deep Neural Networks for Radiological Imaging —
• SurgCast: Action-Conditioned Future Skeletons for Controllable Surgical World Models —
• Sparse All-Layer Connector for Domain Generalised Semantic Segmentation —
• Feedback as an Interface: Controlling Form and Detail in LLM Translation Refinement —
• A Persian Vision-Language-Action Dataset and Benchmark for Ambiguity- and Infeasibility-Aware Instruction-to-Plan Grounding —
• Rate-Resolved Text Evaluation with Masked Diffusion Language Models —
• ProtoPur: Prototype-Purity Pseudo-Labeling for Weakly Supervised Instance Learning in Pathology —
• The Mathematical Collective Unconscious: Jungian Archetypes within the Latent Space —
• Rethinking Human Intent-to-CAD: Parametric CAD Model Generation via Cooperative Multi-Task Alignment and Spatial-Aware Reinforcement Learning —
• Towards Autopilot Labeling for Short Video Content Governance —
• When do explicit penalties improve flow matching? —
• Notes on Uniform Policy Score Bounds and Lipschitz Continuity for Truncated Gaussian, Product Cauchy, Boltzmann, and Softmax Policies —
• Do LLM Tutors Actually Teach? A Process-Centered Synthesis of the Tutoring Loop —
• bharatSetu: a 1M x 22 bridge for indic translation —
• Generalizable Model Editing Could Be More Than Direct Generalization —
• Preserving Updates in Low-Bit Recurrent State Quantization —
• TL;QR: Querying Temporal Discursive Patterns in Textual Corpora —
• To Cache or Not to Cache: Understanding and Mitigating Safety Incompatibility in Cross-Model KV Cache Reuse —
• PhySimCode: A Benchmark and Evaluation Method for Physics Video to Code Generation —
• Visually Legible but Adversarial: Watermark-Constrained Attacks on End-to-End VLMs for Document Understanding —
• NepPlural: A Human-Anchored Benchmark for Persona Annotation in Nepali Migration Discourse —
• ScenePlacer: Reformulating Scene-Level Object Pose Estimation as Visibility Field Learning —
• RSAJEV: Training-Free Abstaining Prototype Memory for Eliminating Silent Hallucination in JEV-Based Remote Sensing Classification —
• SpikeOcc: Towards Resilient Multimodal 3D Occupancy Perception via Spiking Experts —
• Frozen History: When Temporal Losses Stop Accumulation in Learned Real-Time Supersampling —
• Enhancing Zero-Shot Domain Transfer in Low-Resource Sentiment Analysis: A Case Study in Marathi —
• Robust Marathi Sentiment Analysis using Domain-Adaptive Pre-training and Data Augmentation Approaches —
• SlotAVS: Object-Centric Audio-Visual Segmentation via Bi-Modal Slot Attention —
• RSIPO: Recursive Self-Improving Policy Optimization for Agentic Reinforcement Learning via Skill-Policy Co-Evolution —
• Who Sets the Pace of Judge Hacking? Judge Capability versus the Policy-Judge Gap —
• Heterogeneous Structured Pruning Under Resource Constraints for Hardware-Efficient Large Language Models —
• Data-Centric Recursive Self-Improvement for Navigation via Policy-Synchronized Language-Action Dual Bootstrapping —
• Agent Sociodrama: Social-Dynamics-Guided Persona Recasting for LLM Agent Teams —
• Inference on the Value of Policy Improvement in Adaptive Experiments —
• An Explainable System for Offensive Content Detection in Nastaliq Urdu: Corpus Curation, Transformer Benchmarking, and Word-Level Attribution —
• Reference-Efficient Multimodal Bitstream-Corrupted Video Recovery via Depth-Conditioned Adaptation —
• Totem: Object-Centric 3D Memory for World-Consistent Video Generation —
• Adaptive Look, Flexible Thought: Learning How and How Long to Think in Multimodal Reasoning —
• ChangeT cubed: Evaluating Vision Test-Time Training for Remote Sensing Change Detection —
• ML-CLAPPS: A Probabilistic Framework for Imbalanced Multilabel Learning under Class Prioritization —
• Latent Taxonomies: How Property Induction Develops in Language Model Pre-training —
• Binding Phoneme Identity to Position in Neural Models and the Human Brain —
• Mobility Helps Learning: Unsupervised Model Adaptation for Object Recognition via Movement —
• What Do Sparse Autoencoder Feature Discovery Scores Identify? From Activations to Interventions —
• Sharper Frames, Worse Rooms: Viewpoint Coverage Dominates Image Quality in Feed-Forward 3D Reconstruction —
• Unchecked or Falsely Accused? Evaluating AI-Generated Text Detectors on Kazakh —
• RetryBench: Separating Retry Permission from Recovery Outcomes in GPT Tool Use —
• Learning Multimodal Sensor Fusion with JEPA Representations for Breathing Activity and Pose Recognition —
• Minimax Rates and Improved Spectral Convergence for Multi-Manifold Clustering —
• Review Rarely Beats the Stronger Model: Separating Verification from Capability Transfer in LLM Pipelines —
• Gen Through the Lens: Camera-Controllable Video Generation via Cross-Coordinate Conditioning Fusion —
• RIVO: Lightweight Adaptive Rerouting for Deep Research Agents —
• More Reasoning Is Not Always Better: Executing Clinical Guidelines with LLMs and Rules —
• VIEWCRAFT: Evolving What Multimodal Agents See and How They Read It —
• Frequency-Step Cache: Separating Synchronization and Repair for Efficient Autoregressive Video Diffusion —
• When Trackers Fail: VLM-Guided Verification and Recovery for Robust Video Object Segmentation —
• The Anatomy of a Blind Spot: What an LLM Judge Must Be Shown, and How Much —
• Natural Language Processing for Aligning Readings with Learning Objectives and Assessing Topic Balance in Dental Education —
• Unfair and Unsafe: Everyday Bias Triggers Emergent Misalignment —
• Scaling Repeated-Data Pretraining: Batch Size, Regularization, and GPU-Time Tradeoffs —
• Mixed Feelings, Single Polarity —
• Revocation Is Not Complete at Receipt: An Executable Outcome Contract for Distributed Agent-Tool Gateways —
• Representational Difference Distillation —
• PAIR: Memory-Efficient and Fast Inference with Reading--Reasoning Fusion —
• Improving the Evaluation of Biomedical Question Answering with Few Human Preference Signals —
• Meaningful Moments: A Corpus of VLM-Derived Temporal Importance Labels for Video Action Recognition —
• Removing the Compute Bottleneck for Event-based Visual Place Recognition with Spiking Neural Networks —
• Of Errors and Echoes: Responsibility-Aware Credit Assignment for Tool-Use Agents —
• Limit Theory for Online Expectile Estimation under Short and Long Range Dependence —
• A Nonlinear Time-Series Theory for Constant-step SGD with Momentum —
• Beyond the greenwash: do LLMs' ecologically sound decisions hold up under pressure? —
• Backward Conformal Prediction under Covariate Shift —
• From Failure Prediction to Component Selection in Semantic Segmentation —
• PIGEONS: Incremental Long-Video Captioning with Persistent Identity Gallery and Evolving Overarching Narrative Summary —
• Dream2Discern: Learning from Hard Latent Variants for Runtime Failure Prediction in Multi-Agent Systems —
• Regime-Aware Router Learning for Multi-Domain Medical Image Generation with Diffusion Experts —
• GUIDE: Computationally Augmented Consensus for Trustworthy Clinical Practice Guidelines —
• Elephant in the Fridge: Constant-Memory Frame Packing for Long Video Understanding —
• Scaling Multi-Agent Intelligence with Substrate-Centric Coordination —
• Geometry-Guided Semantic Reconstruction for 3D Instance Segmentation —
• A Concept-informed Interpretable Artificial Intelligence Framework for Breast Ultrasound Diagnosis —
• PID-Gated Continual Prompt Adaptation Under Noisy Feedback —
• Query Your Way Out: Benchmarking Multi-Turn Knowledge Reasoning through Generalized Puzzle Solving —
• ProMoDD: Dataset Distillation for 3D Human Motion via Prototype-Conditioned Diffusion Models —
• Omni-Modal LLMs Know More Than They Say —
• Phantom Merge: When Your Large Language Model Agents Pick One but Tell You About Another —
• Distributional Regression for Arbitrary Response Variables Using ReLU Neural Networks —
• All-St3R SLAM: Real-time RGB Dynamic SLAM with Pixel-wise Motion Masking —
• Unravel: Recovering an Interactive 3D Scene from a Single Image —
• Equilibrium Learning in Unbounded Stochastic Games with Gaussian Transitions —
• Think Before You Speak: Enabling Training-Free Latent Communication Between Language Models —
• Esports World Model Dataset —
• Geometric Parrots: Rule Absence vs. Rule Fragility in Spatial and Physical Reasoning for Vision-Language Models —
• DeepThink: A Conceptual Model for Knowledge-Addressable Recurrent Language Modeling —
• EventLens: Event-Structure Reinforcement Learning for Video Understanding —
• EgoAlign: Situated 3D Understanding without Situation Descriptions —
• DIGS: Distribution-Informed Gaussian Splatting for Training-Free Open-Vocabulary 3D Segmentation —
• Exploration strategies for Gaussian process bandit optimization with inference regret guarantees —
• L"ushi-1840: Building and Probing a Knowledge Cutoff for a Classical Chinese Language Model —
• VGGA: Geometry-Grounded Autoencoder for Video Generation —
• MP-MNCA: Morphology-Prior Masked Neighbor Cross-Attention for Efficient Spatial Domain Identification —
• Look Back to Move Forward: Learning to Revisit for Agentic Compositional Image Generation —
• AutoNovo: Agentic Program Search Discovers Transferable Nucleic Acid Design Algorithms —
• Seg2Click: Learning Pixel-Level Action Masks for GUI Visual Grounding —
• From Simulated Fluids to Living Tissue: Zero-Shot Foundation Model Transfer to Perturbed Embryos —
• Avoiding What Your Teacher Already Knows: Entity Familiarity for Deep-Search Distillation —
• MFlowHOI: Whole-Body Human-Object Interaction Generation via Part-Aware MeanFlow —
• EgoWholeView: Camera-Only Head-Mounted Capture of Body and Bimanual Hand Motion for Embodied Demonstrations —
• Hardware-Friendly Token-Group Activation Quantization for Low-Bit Mamba Super-Resolution —
• Error Correction in Trained TTT Layers —
• Graph3R: Unifying View Graph Initialization with Feed-Forward Reconstruction —
• Deep Speech Representations Track Word Timing along a Low-Dimensional Circular Trajectory —
• FlowAvatar: Real-Time Full-Body Avatars from Sparse Egocentric Inputs on Consumer XR Devices —
• DepthTile: Depth-Guided Existing-Image Self-Tiling with Input-Side Tileability Assessment —
• agent-risk: Risk Assessment for Changes to AI Agents —
• Moment-Based Anomaly Detection in High-Dimensional Mixed Data —
• ProGAN-Based Synthetic Image Generation for Industrial Visual Inspection of Metallic Parts —
• REVEAL: Reverse Engineering Agentic Latents from Frontier-Model Traces —
• 3D Gaussian Splatting for Fine-grained Urban Scene Reconstruction via Semantic-guided Density Control and Loss Design —
• ClothAtlas: Joint Garment Identification and State Estimation under Self-Occlusion —
• Learning Temporal Evidence Selection for Video Anomaly Understanding —
• RAR: Rate-Adaptive Rollout for Time-Consistent Interactive World Models —
• Strong yet Controllable LLM Abstention without Learning to Abstain —
• Building age estimation based on widely available multi-source1 geospatial data —
• Anomaly-Guided Observation and Two-Level Semantic Feedback for Training-Free Video Anomaly Analysis —
• Explicitly Aligned Multi-View Breast Cancer Risk Prediction from Variable-Length Screening Histories —
• BAT3R: Robust Online 3D Reconstruction with Bayesian Adaptive State Updates —
• WorldRoPE: Probabilistic World-Centric Rotary Embedding for Multi-View Attention —
• SyncReward: Human-Aligned Synchronization Reward for Audio-Visual Generation —
• RIFT: Token Influence Follows the Path of Least Resistance —
• A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes —
• Self-Distillation with Verbal Regularization —
• VideoSlate: Stateful Commit Control for Temporally Grounded Agentic Video Reasoning —
• HexaBee-4B: Six-Language Translation with Multilingual Answering and Bidirectional Supervision —
• From Sound to Sense: An Anatomy of Semantic Supervision for Speech-Driven Visual Grounding in Remote Sensing Imagery —
• Rule vs.Consequence: Dissociable Internal Representations of Moral Reasoning in LLMs —
• PersonaFactor: Preference Factors with Dynamic Selection for LLM Personalization —
• Throwing4: Phase-Aware Pose Modeling for Throw-Distance Prediction in Elite Athletics —
• RetroPlanner: Multi-Route Expansion and Multi-Objective Reasoning for LLM-Based Retrosynthesis Planning —
• Recovering from Pose-Selection Failures in Repeated Visual Localization —
• Not All Challenges Are Equal: Evaluating Interactive Liveness Detection Against Real-Time Face Swaps —
• Learning omega-Powered Conditional Distributions with Importance-Weighted Flow Matching —
• CultureBot: A Framework for Retrieval-Augmented Exploration of Cultural Heritage Data —
• ARMED: Adaptive Retrieval-Augmented Medication Recommendation —
• VisExMEM: Visual Evidence-Based Explainable Multimodal Evaluation Metric —
• Pixels or Code? Benchmarking Instruction-Guided Visual Artifact Operation —
• Explaining Heterogeneous Treatment Effects with Causal Forests and the Hoeffding Functional Decomposition —
• Tikhonov-Stabilized Bezier Representation Forecasting for Training-free Diffusion Acceleration —
• CivWorldBench: Evaluating Language World Models in Structured Multi-Agent Sandbox Worlds —
• Not Every Cue Is Evidence: Cue Supportability for Forensic Distillation in Face Forgery Detection —
• GHOST: Geometry-Constrained Hybrid Diffusion for Open-Vocabulary 3D Scene Graph Generation —
• The interpreter as developer: An autoethnographic design study of building a computer-assisted interpreting tool through human-LLM co-creation —
• Online Individualized Treatment Selection under Partial Feedback: False Discovery Control and Minimax Optimality —
• Teaching Vision-Language Models to Decline: Support Completion Against Mirage —
• PATMT2: Two-stage domain adaptation for multilingual patent translation —
• Auditing Observation-Quality Shortcuts in the Evaluation of Person-Centered Motion for Video Anomaly Detection —
• Ontology-Based Post-Coordinated Expression Assembly: from Clinical Narrative to Queryable Knowledge Graphs —
• XAI-Guided Fourier Domain Generalisation for Pure-Simulation Sim-to-Real Deep Reinforcement Learning —
• MiMaAlign: Hierarchical Micro-Macro Alignment for Physically Plausible Video Generation —
• Communication, Vocalisation, and Language —
• Self-Verifier Capture in Unified Visual Generators: Diagnosing and Repairing Closed-Loop Self-Reward —
• KARAMEL: Knowledge-Aware Ranking for Multimodal Entity Linking in the Arts via Heterogeneous Graphs —
• Intent-Triggered Anchoring for Multi-Token Format Units in Parallel Decoding of Diffusion Language Models —
• Asymptotic Risk of Learning-Rate Schedules for Least-Squares Regression with Discretely Distributed Features —
• EAU: Environment-Agent-User Co-evolution for Training User-Centric Agents —
• Deep Learning-based Mechanical Characterization of Synthetically Generated Plant Fibers under Transverse Compression —
• Support or Magnitude? Characterizing Neuron Relevance for Instruction Data Selection —
• Spectral-Driven Adaptive-Rank Optimizer for Large Language Model Reinforcement Learning with Verifiable Rewards —
• Learning Graded Sparse Mechanisms: A Representation Theorem and Certified Generalization —
• PatTransBench: an ipc-stratified, multi-dimensional benchmark for patent machine translation evaluation —
• Heuristics are Not Enough: Calibrated and Corrected Caching for Video Diffusion Acceleration —
• Know Yourself and Know Your Neighbour: A Syntactically Informed Self-Supervised Compositional Sentence Representation Learning Framework using a Recursive Hypernetwork —
• NoduleFlow: Automatic 3D Lung Nodule Detection and Segmentation —
• G squared WAM: World Action Models for Efficient and Geometry Grounded Image-Goal Navigation —
• VLA-GAUGE: Systematically Evaluating Generalization and Robustness in Vision-Language-Action Models —
• Dense Endoluminal Mapping and Localisation from Monocular Bronchoscopy —
• NFDiag: High-dimensional MCMC Convergence Diagnosis with Normalizing Flows —
• CineRecon: Benchmarking Frontier Models for Agentic Cinematic Reconstruction —
• Variational Filtering for Bayesian Last Layers with an Independent Binary Probit Surrogate —
• Code-Interactive Visual Reasoning for Financial Candlestick Understanding —
• JtT: Rethinking Point Trajectory Diffusion for Diverse and Rigid Motion from an Image —
• DEFlow: Deep Equilibrium Flow Maximum-Entropy Reinforcement Learning —
• Detecting Chain-of-Thought Hijacking in Backdoored Multimodal Large Reasoning Models —
• MixLID: Token-Level Language Identification in Mixed-Language Text —
• DR-APO: Decoupled-Representation Advantage-Guided Policy Optimization from Continuous Control to Embodied Navigation —
• Answer It or Not? Towards Prompt-Adapted Adversarial Images on Vision Language Models —
• Adaptive Interaction Network for Human Motion Prediction During Human-Robot Collaboration —
• REEL: Learning to Retrieve, Evaluate, and Evolve Visual Memory for Any Long-Horizon Multi-Shot Video Generation —
• SuPER-Mamba: Surprisal Prioritization for Extended-Context Routing in Mamba —
• SIREN: Situation-aware Index for Response and Emergency ActioNability in Crisis Translation —
• Does a Hosted Decision Model Respect Irrelevant Alternatives? A Pre-Registered Test of Option-Set Dependence in Jev —
• Revisiting Natural Language Prompts from Large Language Models for Open-Vocabulary Video Recognition —
• Amortized Variational Inference of Latent Intensities from Replicated Point Patterns —
• EvoSpatial: Experience-Driven Skill and Tool Evolution for Spatial Reasoning Agents —
• EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI —
• 3D-VITAL: Visibility-aware Identity Training from Artificial Liftings of 3D Reconstruction model for Cross-View Object Re-Identification —
• Attention-Graph Curvature for the Verification of Mathematical Reasoning —
• Audit Before You Distill: A Controlled Workflow for Reasoning Transfer into Compact Medical Language Models —
• What Is the "Self" of a Large Language Model? —
• Learn from Move: the Next Step for GUI Agents —
• Lost in Contrast: Instruction-Preserving Decoding for Reliable Visual Question Answering —
• PathPilot: An Adaptive Question-Conditioned Multi-Magnification Navigation Agent for Whole-Slide Image Analysis —
• VAttnPO: Visual Attention-Guided Policy Optimization —
• Amortised adaptive importance sampling —
• FoveaTree: Information-Density Trees for Query-Guided Visual Token Pruning in Video Large Language Models —
• The Midas Touch for Metric Depth: Interpretable Metricization for Generalizable Depth Perception —
• The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language? —
• Is the Platform Part of the Measurement? A Protocol and Simulation Study for Cross-Platform Equivalence of Machine-Labelled Policy Sentiment —
• Image Tokenizers Should Be Trained Between Images —
• Towards Complementary Keypoint Detection via Mixture of Detectors —
• A Bethe-Ansatz Source of Early Token Correlations: A marginal-preserving construction for parallel proposals —
• InfoVis: Unified Slide Image Generation and Editing —
• ViewStruct: View-Grounded Articulation and Connectivity Inference —
• TPPO: Tracking-Value PPO For Language Model Reasoning —
• One-Layer Transformers Provably Learn In-Context K-Nearest Neighbor Prediction with Chain-of-Thought —
• When Repeated Measurements Hurt Tabular Foundation Models —
• X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray Reasoning —
• Carve VFM: Carving Specialists out of Generalists, Automatic Task-Specific Compression of Vision Foundation Models —
• On the Calibration of Crowdsourcing Aggregators —
• Growing to See: Efficient Pretraining of Self-Supervised Vision Transformers —
• Sample-Size Extrapolation in Simulation-Based Inference via Statistical Scaling —
• The Bell Boundary of Shared Representations —
• A Scaled-Up Empirical Study of Syntactic Language Models —
• Label Collapse in Diffusion Language Model In-Context Learning —
• Feature Normalization under Finite Training Budgets —
• A Multi-Task Deep Learning Framework for Pose-Invariant Face Recognition and Facial Landmark Localization —
• Radial Tail Repair: Post-Training Correction for Diffusion and Flow Models —
• Towards Cost-Efficient Video Dataset Construction with Large Language Models —
• ANCHOR: Assessing Naturalistic Constraint Hierarchies and Ordering in Repository-Grounded Agentic Coding —
• Vision Models Capture Complementary Representations in Children That Converge in Adults —
• From Volpiano to Pixels: 30,000 Lines of OMR Ground Truth for Medieval Chant —
• Davinci-Tool: Scaling Agentic Tool-Use Supervision with Verifiable Executable Environments —
• SpecGuardBench: Can Coding Agents Construct Reliable Validation Harnesses from Natural Language Specifications? —
• EviStress: Evidence Stress Learning for Generalizable Face Forgery Detection with Textual Explanations —
• Acceptable Misalignment —
• Barycentric Guidance: Turning Foundation Image Editors into Continuous Affective Controllers —
• Observe, Share, or Predict? Disentangling Partner Information in Embodied Multi-Agent Collaboration —
• Mining Cross-Modal Perception Gaps in Autonomous Driving: When Camera and LiDAR Disagree —
• ConDenseDrive: Knowledge-Dense, Token-Sparse Visual Encoder for End-to-End Autonomous Driving —
• FOCUS: A Subspecialty Foundation Model for Intraoperative Central Nervous System Pathology —
• Minimax Optimality of Shampoo for Nonstationary Stochastic Optimization —
• Estimating Uncertainty of Omnimodal Large Language Models via External Omnimodal Embeddings —
• UniMetric: Universal Metric Scale Recovery for Visual Geometry Foundation Models —
• Lexical Redefinition Reveals Context-Sensitive Instruction Following —
• DDVR-Net: Degradation-Aware Decoupling and Visibility-Guided Restoration for Historical Document Recognition —
• SarcasTraps: Effect of Sarcasm on Turkish Hate Speech Detection —
• Proactive Physical AI through Continuous Observation and Timely Intervention —
• Event Memory and Narrative Dynamics: A Large Language Model Framework for Multidimensional State Spaces —
• The Value of Being Wrong: Self-Mined Visual In-Context Learning from Errors —
• Predicting Mental Health Help-Seeking Behaviour on Reddit: A Survival Analysis of Linguistic and Behavioural Signals —
• AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding —
• Self-Supervised Learning of Semantically Consistent Visual Features via a Simple 3D Prior —
• MaskSense: Confronting the Visual Exploration Trap in Masked Image Generation —
• MAGNET: Manifold-Aware Graph Diffusion Network for Connectome Generation —
• DIME: Scaling Facial Representation Learning via Differential Masked Autoencoding —
• ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty —
• Understanding Personalization Gains in Whisper —
• MENTOR: Procedural Shielding of Tool-Using LLM Agents —
• Beyond Model Bias: Cross-Entropy Based Rankings in Literary and Academic Texts Remain Stable across Small Language Models —
• AD4EDU: Benchmarking University Lecture Audio Description across Accessibility Standards, Metrics, and Learner Preferences —
• Measuring Variable Importance Using Proportional Inclusive Marginal Effects —
• VL-DocIR: A Benchmark for Vision-Based Long Document Information Retrieval —
• In-context classification: Calibration and noise in linear attention with softmax readout —
• PerturbBridge: Direct Phenotypic Supervision for Cellular Perturbation Generation —
• A unified three-branch deep learning framework for automated detection and counting of wildlife in aerial imagery —
• UZHCL for the win at Touche: Encoder Fine-Tuning, Synthetic Augmentation, and an Underpowered Encoder-LLM Comparison —
• Rethinking LLM-Guided Visual Question Answering: Does Deeper Knowledge Integration Really Help? —
• Scaling Unified Multimodal Models to Long-Horizon Interleaved Generation —
• Towards Universal Calibration for In-Context Classification —
• Contrastive learning for video reuse detection —
• Calibrated Typed Decisions for Entity, Relation, and Event Extraction —
• Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations —
• When LLM Agents See a Conflict but Still Proceed: Cross-Domain Dependencies in Network Slicing —
• Can R'enyi Accountants Beat Lossless Accountants for DP-SGD? —
• D2: High-Capability Full-Duplex Spoken Language Models from Half-Duplex Backbones —
• ToolHACE: A Dataset for Detecting Hallucinations in Tool-Based LLM Responses —
• EgoProbe: Benchmarking Long-Range and Exact Memory in Streaming Egocentric Videos —
• MARS: Systematic Multi-Level Acceleration for World Action Models —
• SCOPE Before You Sample: Conformal Scores Guide Posterior Exploration for Fast and Calibrated Bayesian Predictive Inference —
• The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express —
• A Generalized McDiarmid Inequality: Sample Reuse and Non-Uniform Unbounded Differences —
• Learning Perceptual Geometry of Natural Odor Mixtures from LLM-Derived Descriptors —
• ChemExplain: A Multimodal Benchmark for Explanatory Chemical Reasoning —
• DuoDiscover: Scaling Autonomous Discovery Beyond the Evaluation Bottleneck —
• Soft Teacher Forcing —
• Welcome to the Age of Agent Ads: Using On-Policy Attacks to Steer Agentic Decisions —
• Multics: Generative Multi-Player World Model with Hierarchical Communication —
• MLAfication: Efficiently Enabling High-Fidelity MLA for Modern LLMs —
• Online LLM Personalization from Pairwise Preferences Across Users —
• AReA: Adaptive Receptive-Field Attention for Efficient High-Resolution Diffusion Transformers —
• Beyond Sparse Captions: Aligning Slide-Level Text and Patch-Level Vision in Pathology —
• IsoVRBench: Video Reward Modeling Without Human Feedback —
• When Left Is Right? Camera-Motion Understanding beyond Screen —
• From Conceptual EneMonIE to an Executable Backend: A Modular Architecture for Energy Information Extractionand Grounded Question Answering —
• HybridDualPathNet: Cross-Fusion Attention of EfficientNetV2 and Swin Transformer for Brain MRI Classification of Alzheimer's and Parkinson's Disease —
• Nufo: A Benchmark for Ewe Automatic Speech Recognition —
• The Surprising Expressivity of a Single Neuron with Autoregressive Chain-of-Thought —
• StartLux-Decision: AutoThink Decision Models with Reinforcement Learning for Calibrated Deliberative Decisions —
• Beyond Successful Play: Diagnosing Opponent Modeling in Large Language Models —
• Deeply unrolled alternating projections for phase retrieval —
• From Sampled Evidence to Radiology Reports: Objectives, Constraints, and Realization —
• Near-Optimal Regret Bounds for Monotone Stochastic Optimization under Semi-Bandit Feedback —
• SSM-Attack: Learning Where to Perturb for Black-Box Face Privacy —
• Thinking Outside the Bots: Human Ideas in AI's Semantic Blind Spots Are Not Inferior —
• The Curse of Knowledge in Role-Playing Language Models —
• COMPASS: Membership Inference Attack Against Vision Large Language Models via Adaptive Noise Calibration —
• Accounting for model error in ensemble variational data assimilation through learned generative dynamics —
• Monitoring for Shifts at Unknown Times: Parameter-Free Conditional Conformal Martingales —
• On the Statistical Mechanics of Attention —
• HairFlow: Modeling 3D Hair Strands with Multi-Modal Conditioned Diffusion —
• MID: Mask-Image Distributional Divergence for Evaluating Medical Image Segmentation —
• WIRE: Weight Space Implicit Representation Encoding of Frozen Video LLMs —
• Brain2Face: A Study on the Relation of EEG Signals and Spatio-temporal Head Avatars —
• claimcheck: Measuring and Improving Specification Faithfulness —
• UMCR: Universal Medical Concept Recognition through Guideline Integration and Agentic Linking —
• PIG: Parallel Inconsistency Guidance for Accelerating Diffusion Language Models —
• What Makes a Latent Space Shared? Direct and Recoverable Compatibility in Motion VAEs —
• EPOCH: Coverage-Aware Evolution with Experiential Preference Optimization for LLM-based Automated Heuristic Design —
• Mitigating Black-Box Score-Based Query Attacks via Gradient Confusion —
• Measure Neural Mappings —
• Learning to Detect Shifts in Unstructured Data Streams via Monitoring-by-Betting —
• Evaluating Temporal Reasoning in Large Language Models: Event Temporal Relation Extraction with Allen's Interval Algebra, Logical Consistency, and Relative Time in Narratives —
• Bottlenecked semantic compression by representation reconstruction —
• World-T3: Test-Time Training as Persistent Memory for Real-Time Video World Models —
• Know-Thy-Not: Training-Free Targeted Negation in Dual-Encoder Vision-Language Models —
• The belief-action gap in language model agents —
• TableVision: Recovering Verified Structure from Historical Tables —
• Effect of Spherical Parameter Space on Number of Non-Forgetting Tasks and Capacity of Continual Learning —
• Compositional RL: Evaluating and Training LLMs for Collaborative Reasoning —
• Distilling Text LMs into Speech-to-Speech LMs by Matching Internal Text Predictions —
• Efficient Fr'echet Distributional Decoder Alignment for Latent Generative Models —
• A Sense of Proprioception in Language Models —
• C NL: Collapse-Based Active Neural Learning with Objective-Aligned Representation Control —
• Gaussian Splatting for Real-Time Scene Reconstruction: Methods, Systems, and Trade-offs —
• TwinPrune: Density-Aware Two-Phase Token Pruning for Vision-Language Models —
• CrossPortrait: Cross-Species Portrait Animation via Adversarial Motion Alignment —
• Everyone Was Right: Lexical Representations in Language Models Are Both Abstract and Holistic —
• TacitVLA: Reasoning Vision-Language-Action Models Without Reasoning Latency —
• Many Steps into One: Reasoning Contraction and the Limits of Safe Memory Reuse for LLM Agents —
• Reference Wording Shapes Retrieval of Robot Task Instructions —
• FlashVSR+: Efficient Video Super-Resolution via Hybrid Attention and Reward-Yield Distribution Matching —
• EdgeWise: Bringing Long-Horizon Agents to the Edge via Selective Cloud Delegation —
• Benchmarking Self-Supervised 3D Medical Vision Foundation Models and Transferability Estimation —
• Dissociating Answerability from Extraction in LLM Question Answering over Radiology Reports —
• Risk-Averse Inference-Time Diffusion Model Alignment with Finite-Sample Guarantees —
• Physical-Field-Bridged Material Parameter Inference from Multi-View Deformation Videos —
• Tanawpamana: Benchmarking Vision-Language Models on Philippine Cultural Heritage Sites —
• TDR: Tagalog Diacritic Restoration Dataset —
• Morphology-aware deep learning predicts colorectal cancer survival from whole-slide images —
• PLEDGE: Progressive Lossy Error-Bounded Compression of 3D Gaussian Splatting —
• Facts Don't Speak Louder than Words: The Behavioral Essence of Long-CoT Distillation —
• When Does an ASR Error Matter? Task-Role Diagnostics for Tool-Using Voice Agents —
• Smartphone-Based, Field-Deployed AI Inspection System for Power Distribution Infrastructure —
• RELAY: Selective Self-Distillation with Adaptive Hand-off for Rubric-Based Reinforcement Learning —
• 4LC: 4D Latent Cache for Efficient Conditioning in Dynamic World Modeling —
• Executable Prefix Values for Locating Agent Failures —
• Calibrated Disambiguation for Partial Multi-label Learning —
• FrameLite: Decision-Oriented Aesthetic Cropping with a Compact Vision-Language Model —
• MemAct: Streaming Spatial Understanding with Memory and Active Exploration —
• DisaggKD: Efficient Knowledge Distillation through Teacher-Student Disaggregation and a Fused Loss —
• Decouple to Forget: Semantic Mediation for Sensitive Unlearning in Large Language Models —
• textGrain watermarking is harder to detect: can sum-tests still do it optimally? —
• Open-o3-Video-v2: Seek, Ground, and Reason with Verifiable Spatio-Temporal Evidence —
• Worth Their Weights? Evaluating Generative Models for Discourse Parsing —
• ScenePercept: Scenario-Conditioned Perceptual Assessment of Real-World User-Generated Content Images —
• The Image Still Has a Say: Final-Answer Repair for Multimodal Chain-of-Thought —
• DeltaTTG: Training Multi-Turn Agents with a Token-to-Go Progress Meter —
• Hierarchical Curriculum Learning for Surgical Video Generation: Focusing on Bilateral Axillo-Breast Approach Robotic Thyroidectomy Datasets —
• LatentRiskBench: Measuring State-Dependent Risks in Enterprise Agents —
• Encoding Context into LLM Weights: What Makes It Context? —
• MedEvo: Medical Reasoning Evolution through Knowledge Navigation —
• SWORD: Scene-Grounded Wasserstein Object-Relation Distance for 3D Scene Graph Prediction —
• Understanding Gene Expression-to-Image Learning: Spatial Structure and Information Fidelity —
• Pushback in a Haystack: Language Models Respond to Factual Errors Ambiguously and Ineffectively —
• A Differentiable 3D Scene Graph Metric via Contextual Hellinger Triplet Geometry —
• From Token Correction to Claim Admission: Training-Free Hallucination Mitigation for Large Vision-Language Models —
• Where Small Mistakes Matter: Quantifying State Criticality for Imitation Learning —
• ECG-RSI: Recursive Self-Improvement through Targeted Electrocardiogram Waveform Evolution —
• DT-LoRA: Reaching Beyond the Tangent Space of Low-Rank Adaptation —
• Understanding Machine Taste: Aesthetic Agreement Across Foundation Models in K-pop —
• Geometric Structure Predicts Per-Example Neighborhood Loss Under PCA Truncation —
• AttenIR: Aligning Dense Visual Attention with Mesh Warping for Unsupervised Image Retargeting —
• Auditing Failure Attribution: Label Predictability and Judge Behavior —
• Gaussian Process Bandits for AI System Optimization with Selectively Verified Feedback —
• Connecting the Dots: Enduring Performance Gaps in ASR Systems Between Standard and Minority English Varieties —
• Autonomous 3D Visual-Thermal Inspection for Utility-Scale Solar Farms —
• Efficiently Learning To Reason or Not to Reason: Root-token Policy Optimization for Adaptive Thinking —
• Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders —
• Beyond Belief Nesting: Rethinking Higher-Order Theory of Mind in Large Language Models —
• GENDIS: Gender-Related Discourse and Stance in Multilingual YouTube Comments from India, Kenya, and Nigeria —
• Full-duplex Spoken Language Modeling with Discrete Diffusion —
• Where Words Begin and End —
• When, Where, and Why: Hierarchical Causal Explainability for Video Face Understanding with MLLMs —
• MHVF: Training-Free GUI Grounding via Verification-Guided Multi-Hypothesis Search —
• Multimodal Fusion of RGB and Complementary Modalities for Semantic Segmentation —
• BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems —
• SPARCL: Spatially Adaptive Regional Coverage for Active Learning in Chest X-ray Analysis —
• Perturbation, Not Provenance: Why Activation Steering Does Not Make a Robust Text Watermark —
• Follow the Perturbed Leader for Linear Bandits: sqrt T Regret over General Convex Sets —
• Checking the Checkers: A Synthesis of Verifier Noise and Robustness in RLVR —
• QK-Edit: Revisiting Attention-based Injection in MM-DiT for Image and Video Editing —
• What Makes a Good Visual Representation for Spatial Understanding in MLLMs? —
• DVTR-KD: Distilling Vision-Language Models via Intermediate Visual Token Reconstruction —
• Frequency Routing and Identity Conditioning for Thermal-to-Visible Face Translation —
• MicroCLIP: Connecting Text and Microscopy Images via Contrastive Language--Image Pretraining —
• Prefilling-dLLM v2: Learned Chunk Routing for Long-Context Diffusion Language Models —
• LoopREG: Representation Supervision Enables Test-Time Recurrent Depth in Diffusion Transformers —
• Cross-Lingual Transfer of Evil Vectors Is Directionally Asymmetric —
• Diagnose-to-Generate: Eliminating Structural Distortions in Flow Models via Patch-Level Grounded Alignment —
• Training LLMs for Multi-Agent Collaboration with Verifiable Information Rewards —
• Swiss4Arg: A Parallel Dataset for Multilingual Argument Mining across Swiss National Languages —
• Evaluating the Causal Faithfulness of SHAP as a Candidate Attribution Method for SAE Features —
• Visual Dependence Does Not Imply Answerability: A Systematic Analysis of Answerability Signals in Frozen InstructBLIP —
• DriveWLA: Unifying Future Dynamics and Scene Semantics through World-Language-Action Modeling for Autonomous Driving —
• Text-Only vs. Image-Aware VLM Judges for Manga Translation Evaluation —
• LinMorph: Learning to Morph Weights for Efficient Video DiT Linearization —
• Batch-Paced: Fixed-Budget Batch Decoding in Discrete Diffusion Language Models —
• CineMem: Attention-Aligned Fine-grained Memory Retrieval for Multi-Shot Long Video Generation —
• HierarchiRetina: A Diabetic Retinopathy Screening and Grading Cascade with Heterogeneous Mixture-of-Experts Lesion Segmentation —
• PUGS: Physics-Guided Uncertainty Modeling for Underwater Dynamic 3D Gaussian Splatting —
• MERIT: Multimodal Evidence Reasoning over Image Tables —
• Sea-Net: A Small and Inherently Interpretable Neural Network for Time Series Classification —
• AI-Enhanced Hyperspectral Imaging for Intraoperative Detection of Low-Grade Glioma: A Clinical Study —
• HaloBlock: Faster and Better Decoding for Diffusion Large Language Models —
• FETALFusion: Resolution-Aware State-Space Modeling for Multi-Domain Fetal Head Segmentation —
• Agent-Reflex: Teaching Language Agents to Act Reflectively —
• Learning Distributions from Samples and Qualities —
• Virtual Head Attention —
• dMRI-F: A Foundation Model for Generalizable Diffusion MRI Analysis from Sparse Acquisitions —
• Web Scale on a Budget: Tail-Aware Retrieval Evaluation —
• Seeing Through the Forgery Is Not Enough: Regressive Sycophancy under Fabricated Evidence in Medical Vision-Language Models —
• PolUR: Polarization Aware Multimodal Degradation Perception for Underwater Image Restoration —
• Counterfactual Certification for Sequential Unlearning with Long-Run Performance Guarantees —
• Uncertainty-Aware Semi-Supervised Framework with Supervised Contrastive Learning for Sugarcane Leaf Disease Detection —
• Saliency-Guided Replay for Continual Learning in Deep Spiking Neural Networks —
• Match the Canvas, Not Just the Answer: Predictability-Aware SFT for Diffusion Language Models —
• Surface Reconstruction from Range Scans Under Visibility Constraints —
• LatentRemedy: Efficient Remedy For Non-Prefix KV Cache Reuse in Multi-Agent Systems —
• Dynamic Interpolation for Training-Time Prediction Reshaping in Low-Light Image Enhancement —
• Machine--human agreement in marking undergraduate assignments is specific to model, prompt and institution —
• FlowSim: Rendering-Free Closed-Loop Simulation for Evaluating End-to-End Driving Systems —
• COMPASS: Contextual Moment Prior Alignment for Generalizable AIGI Detection —
• Weighting-Based Strategies for Classification: Analysis and Local Peak Dynamics with CART Trees —
• Scaling Vision Transformers Without Training: Prune-and-Merge Token Compression —
• inLoRA-MINT: Internal Representations for Auditing Training Data in Domain-adapted LLMs —
• Reliable Scale-Free Bandits via Empirical Likelihood —
• A Benchmark for Bangla Task-Oriented Dialogue Understanding —
• From Agents to Edges: Paired Optimization for Multi-Agent Collaboration —
• LensErase: A Foundation Model for Erase Anything —
• CompanionSafe: From Multi-turn Safety Evaluation to Guardrails for AI Companions —
• APT: Feedback-Driven Agentic Post-Training for Text-to-Image Generation —
• Chain of Thought Monitoring Can Fail Even When CoT Is Necessary —
• SYNPRED: A Synergistic Approach to Multimodal Learning for Clinical Prediction —
• Narrative Lookahead in LLMs: Event-Level Decodability, Predictive Advantage, and Steering —
• Cheap, Consistent, Calibrated, but Biased? Typed Decision Models as Judges —
• H-DDCM: Hadamard Codebooks for Fast Generative Image and Video Compression —
• Monitoring latent reasoning in looped language models —
• Human-Inspired Asymmetric Collaboration for Dual-Student Semi-Supervised Medical Image Segmentation —
• SparPR: Runtime KV-Cache Recovery for Resource-Constrained LLM Inference on Consumer Hardware —
• FlowCred: Hierarchical Credit Assignment and Functional Coordination for Joint Text-to-Video Post-Training —
• Attacking Only Half the Data: Two-Sided Evaluation of AI-Text Detector Robustness —
• Learning a Group Element from Its Action: Galois Correspondences, Optimal ERM, and an Abel-Ruffini Threshold for the Natarajan Dimension —
• Fan Out or Relay? Ownership-Aware Routing for LLM Agents Querying Siloed Tables —
• Claim Framing as a Driver of LLM Sycophancy —
• Benchmarking Multimodal Large Language Models for Zero-Shot Human-Object Interaction —
• LoOPSD: On-Policy Self-Distillation with Looped Teachers —
• Lifting Finite-Sample Gaussian Approximation of Wide Neural Networks to Functional Convergence —
• TRACESTEER: Trace Driven Activation Steering for Hallucination Mitigation in Diffusion Language Models —
• VARI: Robust Text-to-Video Generation via Group-Structured Intent Alignment —
• Hierarchical Time Series Forecasting with Operator-Valued Kernels —
• Closing the Estimation Gap: Training-Free Transductive Inference for Cross-Domain Few-Shot Action Recognition —
• Operator Surrogates —
• TextualRL: Textual Reinforcement Learning for Context Management —
• A Controlled Evaluation of Curvature Inspired Scalar Feature Scaling for Domain Generalization —
• Consider, Then Choose: Order-Invariant Decisions in System One Models —
• Heading Hierarchy Reconstruction for Complex and Long Documents with Level-Wise Multimodal Memory —
• LiFUSE: Frequency-Spatial Unified Self-Supervised Pretraining for Liver Disease Analysis on Non-contrast CT —
• PACT: A Pragmatic Prompting Framework for Expressive and Voice-Preserving Speech-to-Speech Translation —
• Are We Testing Too Much? A Survey of Benchmark Compression for Efficient Evaluation of Large Language Models —
• FALLNEST: Dual-Domain Dataset Condensation of Imbalanced IMU Data for Wearable Fall Detection —
• PAINT: A Framework for Multilingual Pain Representation from Medical Licensing Exams to Real-World Clinical Texts —
• JevGPT: Thinking Fast and Slow in Autonomous Research —
• Traditional and Recent Approaches for Infrared Small Target Detection: Overview, Challenges and Perspectives (Part 1) —
• Efficient Scene-Text Refinement via Pixel-Space Grouped Image Generation —
• LongForcing: Scaling Temporal Manifold Matching for Long-Horizon World Rollout —
• Unsupervised and Supervised Approaches for Infrared Small Target Detection: Overview, Challenges and Perspectives (Part 2) —
• Robust Logical Anomaly Detection under Capture Shifts via Synthetic Anomalies —
• LegalRewardBench: Benchmarking Reward Models for Legal Preference Evaluation —
• AffectWorld: Benchmarking Emotion-Driven Future World Prediction —
• SimInter: Evidence-Grounded User Simulation with Interview-Based Profile Enrichment for Conversational Recommender Systems —
• Prompt or Policy? Optimizing Physician Slate Presentations with LLM User Simulators —
• Omni-RS: Unifying Multi-Task Understanding and Generation in Remote Sensing —
• ShoeFit: A New Dataset and Dual-image-stream DiT Framework for Virtual Footwear Try-On —
• Probing the Geometry of Color Representations in Multilingual Language Models —
• On the Optimality of Random Features under General Conditions: Misspecified Cases and Robust Learning —
• CityCode: Building Agent-Operable 3D Cities from Sparse Aerial Observations —
• Lightweight but Linguistically Fair? Quantisation Robustness across Bavarian, German, and English —
• Urbex: Agentic Spatial Grounding for City-Scale 3D Scenes —
• Hallucination-aware Faithful Reasoning in Remote Sensing —
• SchemaPose: RGB-Based Category-Level Pose Estimation with Parametric Category Schema —
• Noise2Match: Self-Supervised Low-Dose CT Denoising via Adaptive Inter-Slice Matching —
• ReDiS: Reasoning-Diagnosis Self-Evolution for Multimodal Reasoning in Remote Sensing —
• PRISM: Calibrating Black-Box Judges with Counterfactual Probes —
• Bifocal: Training-Free Mixed Granularity Visual Token Pruning for Multimodal LLMs —
• Multi-Stream Residual Computation in Diffusion Language Models —
• Prompt-Injection Probes Learn the Conversation, Not Just the Injection —
• Not All Drift Is a Mechanism: Tracing Prompt Adaptation in LLMs —
• Revisiting the Estimation of Unigram Surprisal in Computational Psycholinguistics —
• Do Factuality Methods Make LLMs More Factual? Reproducibility and Fair Comparison —
• Resource-Efficient 3D Shape and Orientation Estimation of Road Users using Roadside RGB Camera and 3D Radar —
• Agaram 2.5B: The First Tamil-First Language Model Trained from Scratch with Tool Calling. A Playbook for Under-Represented Languages —
• Receiver Decides Agent-to-Agent Negotiations —
• Reliable Memory from Unstable Growth: Curating Self-Written Agent Playbooks with Randomized Experiments —
• Convergence-Rate-Guided Algorithm Design of Expected Improvement for Continuous Simulation Optimization —
• Estimating Surprisal from the Speech Signal —
• Copying Is Not Grounding: Training Faithful Long-Form RAG with Authored-Sentence Credit —
• CoTL: Compact Token Learning for 3D Scene Representation —
• Thinker-WM: A Unified World Model with Shared Multimodal Representations for Embodied Intelligence —
• EmoGenie: Benchmarking and Evaluating Expressed and Evoked Emotion in T2I Generation with MLLMs —
• S cubed-PnP: Plug-and-Play Denoising for Scientific Imaging with Self-Supervised Score Priors —
• Teacher-Guided Tool Learning for Small Language Models via Bidirectional KL Constraints —
• MFVSR: A Multi-Style and Fine-Grained Benchmark for Video Subtitle Removal —
• Readout Scaling Shapes Sparse Feature Recovery —
• Simplifying and Improving On-Policy Distillation for Few-Step Diffusion —
• Physics-Aware Deepfake Detection via Distance-Speech Consistency —
• Recalibrating Many Questions with Few Labels: Minimax Rates and Identification Limits —
• Judges Follow the Reference: The Derivation Gap in Verifying Tool-Call Arguments —
• Aligning Faithfulness for Theory-Scale Autoformalization —
• EgoOmniWorld: Unifying Egocentric Understanding via Ego-Action-World Interactions —
• How Low Can You Go: Fine-Tuning LLMs with Bayesian Optimization (LlMBO) —
• The Last Mile of Coding Agents for Long-Document Tasks: An Empirical Study on Clinical Trial Protocol Digitization —
• Two For The Price Of One: Curvature-Aware Pruning Preserves Influence Functions Too —
• Imprinting: Knowledge Distillation Can Shift Student Behavior —
• Dual-Domain Sparse-View 4D CBCT Reconstruction with Spatial-Temporal Consistency Learning —
• The Geometry of Language Model Logits: Generic Structure and Learned Departures —
• Retrieval Disagreement as a Signal: Diagnosing and Steering Multilingual Dense Embedders —
• How Reliable Are System-One Decisions in Medicine? A Comparative Study of Jev, Jev-like Models, and LLMs —
• Meta-Video: Efficient Standardized Evaluation for Video Understanding —
• Empathy Is Not the Scapegoat for Declines in Factuality and Reliability —
• Beyond the Encoder: Object-Centric Representations Emerging from Segmentation Foundation Models —
• Understanding Circulant Permutation to Extend the CMinHash Estimator —
• Unbalanced W-Flow: Unbalanced Wasserstein Drifting under Finite-Batch Marginal Fluctuations —
• Denoise X: Self-Supervised U-Net for Quantum Mottle Suppression in Chest Radiography —
• Observation Quality in DSM-Assisted Aerial Localization: A Benchmark and Base-Anchored Uncertainty Calibration —
• Rethinking Quality in Large Language Model Routing: Security-Conditioned Model-and-Budget Selection —
• CREATE: Creativity Evaluation of VLMs using the Visual Remote Associates Test —
• Parameter-Efficient Multimodal Brain Tumor Segmentation With Stage-Wise Shared Adapters —
• The Many Morals of AI: The cautionary bias of North American news narratives in global perspective —
• Detecting Unreliable Classifications from Quantized Models —
• Decomposing and Predicting Annotation Disagreement in Harmful Content Across Languages —
• Rethinking Projector Training in Multimodal LLMs —
• AgentStripes: An Evidence-Graded Maturity Model for AI Agents —
• LLM2R: From Large Language Models to 3D Reassemblers —
• From Hindsight to Foresight: Dissecting Clinical Evidence Acquisition in Large Language Models —
• Beyond Answer Voting: Fine-Grained Correction of LLM Reasoning through Graph Alignment across Sampled Traces —
• Recasting DiT for Extreme Image Compression with Hierarchical Visual Priors —
• Jev-ToM: Evaluating Theory of Mind in Typed Decision Models —
• Legal-R1: Agentic Retrieval for Evidence-Based Legal Reasoning —
• Does self-distillation improve reasoning? —
• VideoJev: System One Model for Video Understanding and Interaction —
• Knowledge Graphs as Files: An Agent Harness Where Tool Creation and Tool Use Co-Evolve —
• Two Routes to Interoceptive Prioritisation: Dissociating Likelihood Precision and Preference Reweighting in Active Inference —
• MajorTOM:Elliot Pretrain, A Multi-Sensor, Multi-Temporal Pretraining Dataset for Earth Observation —
• Breakdown of Conversational Involvement: A Joint Multi-Modal Analysis of Conversation Floors and F-formation Detection with Multiple Features —
• To Align or Not To Align: Check Your COMPASS Before You Train —
• ST-GRU PredRNN-V2: A GRU-Based Reformulation of PredRNN-V2 for Video Prediction —
• Does Pretraining Teach Chemical Language Models Chemistry? —
• Texture Interpolation as Texton Matching —
• Bayesian Test-Time Inference of Task-Aligned Similarity from Weak Interactive Feedback —
• BridgeTwist: Twisting Schr"odinger Bridges for Training-Free Conditional Sampling —
• PlanPath: Exploration-Guided Orchestration for Patient-Level Pathology Reasoning —
• Known-Voice Cloning and the Limits of Personality Transfer for Character Design —
• SMI: Ontology-Grounded Semantic IDs for Hierarchy-Aware Medical Concept Representation —
• From Relation-Specific Heads to Shared Interventions: Contextual Entrainment in Language Models —
• Model-Based Online Decision Making via Generative Trajectory Planning —
• RadarGSOcc: Doppler-Aware 3D Semantic Occupancy Prediction Using 4D Radar Tensor with Geometric-Semantic Feature Extraction —
• RSIgen: Reciprocal Self-Improvement for Agentic Visual Generation —
• Just Respect the Semantics: Image-Token Compression via Pre-LLM Semantic Variable-Resolution Tokenization and In-LLM Relevance Pruning —
• Orientation Estimation of Cryo-EM Images for Molecules with Arbitrary Symmetry by Semidefinite Programming —
• GeoReader: Learning Latent Geometry for Spatial Reasoning in Vision-Language Models —
• CryoTracer: A Synthetic-to-Real System for Filament Instance Tracing in Cryo-EM —
• AIRRI: Towards a Human-Readable, Machine-Resistant Reading Interface —
• Code-Switching Detection via Language-Selective Dimensions in Speech Foundation Models —
• On the Monitorability of Natural Language Autoencoders: Identification vs Discrimination —
• Segment Where You Look: Attention as the Interface Between Language Models and Segmenters —
• Neural Block-wise Auto-exposure for Single-Shot High Dynamic Range Imaging —
• Concealed Handgun Detection using WiFi Signals —
• A Rate Transfer Method: Convergence Rates in Stochastic Gradient Descent via Stochastic Calculus —
• Context-aware Bayesian Fusion for Multimodal Object Detection —
• Looking Complete Is Not Being Grounded: Evaluating LLM Sitreps Against Human-Authored Situation Reports —
• LayerCanvas: Infographic Generation in AI Learning Assistants via Hybrid Layer-Set Diffusion —
• Rubrics from the Task, Verdicts from Evidence: Pointwise Rubric Rewards with Tool-Augmented Judges —
• MIST: Multi-Scale Spatial Context for Gene Expression Prediction from Histology —
• The Federated Tumor Segmentation and Classification (FeTS++) Challenge —
• GAD: Geometry Attention Distillation for Consistent Video Generation —
• BLADE: Auditing Information Use for Failure Attribution in LLM Multi-Agent Systems —
• Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agents —
• Where Do LLM Ensemble Gains Come From? Decomposing Candidate Value, Diversity, and Aggregation —
• Trust Before Autonomy: Rethinking Agentic Remote Sensing for Crisis Response —
• Registration is a Representation Problem: Solving Whole-Body PET/CT Registration —
• SkillScaffold: Scaffolding Agentic Reinforcement Learning with Policy-Adaptive Skill Credit Assignment and Selective Injection —
• Self-Evolving Agent Harness —
• Back-Translation Still Matters: Leveraging Monolingual Data for Low-Resource Machine Translation with LLMs —
• DuplexBridge: Coordinating Understanding and Generation for Unified Image Editing —
• Read the Room Before Your Model Generates —
• Joint Neural Posterior Estimation and Model Selection for Chemical Reaction Networks —
• OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding —
• Game2Real-Bench: Do Toy Games Reveal and Transfer to Real-World Visual Reasoning? —
• Plan-Bench: Can Existing VLMs Go Well Beyond Perception to Reason and Plan in Embodied Worlds? —
• Who Improves the Robot? Toward Physical Recursive Self-Improvement —
• The Last Remote Sensing Model Built by Humans? Casting Any Earth Observation Task as Recursive Self-Improvement —
• MFG: Multi-turn Feedback Guided Molecular Optimization —
• Hybrid Reasoning Agents —
• DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning —
• Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards —

Important terms

Diffusion Models for Protein Shrinking

This technique uses diffusion models to physically shrink proteins while ensuring their essential functions remain intact, addressing real-world molecular interaction constraints.

Gene Expression-to-Image Learning

This research examines how the physical layout of a cell influences gene expression patterns and how information fidelity impacts this learning process.

Graph Neural Network (GNN) for Biological Networks

Using GNNs within a GATTACA Framework, researchers are developing methods to control complex biological systems by guiding reinforcement learning decisions.

Active Inference and Interoceptive Prioritisation

This framework attempts to separate how accurately we estimate a belief from how much we value it, helping us understand internal prioritization in the brain.

AeroJEPA

A significant development for aerodynamics, this model learns semantic latent representations to create scalable 3D aerodynamic field models for designing efficient aircraft.