Papers

New on arXiv

Wed 28 Jan 2026

404 papers

Benchmarking Distilled Language Models: Performance and Efficiency in Resource-Constrained SettingsKnowledge distillation → Knowledge distillation for LLMs
ConfSpec: Efficient Step-Level Speculative Reasoning via Confidence-Gated VerificationLLM reasoning → Test-time compute and efficient thinking
Reducing Text Bias in Synthetically Generated MCQAs for VLMs in Autonomous DrivingMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection +1
MARVL: Multi-Stage Guidance for Robotic Manipulation via Vision-Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
WorldVQA: Measuring Atomic World Knowledge in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
From Sparse Decisions to Dense Reasoning: A Multi-attribute Trajectory Paradigm for Multimodal ModerationLLM safety and security → Vision-language and text-to-image safety
HMVLA: Hyperbolic Multimodal Fusion for Vision-Language-Action ModelsVision-language-action models → Generalist VLA models and benchmarks +1
CADENT: Gated Hybrid Distillation for Sample-Efficient Transfer in Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
LLaVA-FA: Learning Fourier Approximation for Compressing Large Multimodal ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
PredictionMarketBench: A SWE-bench-Style Framework for Backtesting Trading Agents on Prediction MarketsLLM agents and tool use → Financial trading and forecasting agents
ALIGN: Aligned Delegation with Performance Guarantees for Multi-Agent LLM ReasoningLLM reasoning → Reasoning for agents, planning and code +1
PhaseCoder: Microphone Geometry-Agnostic Spatial Audio Understanding for Multimodal LLMsMultimodal LLMs / vision-language models → Video understanding with MLLMs
Multi-task Code LLMs: Data Mix or Model Merge?LLM agents and tool use → General LLM agent frameworks and benchmarks
Planner-Auditor Twin: Agentic Discharge Planning with FHIR-Based LLM Planning, Guideline Recall, Optional Caching and Self-ImprovementLLM agents and tool use → Medical and scientific domain agents +1
How does information access affect LLM monitors' ability to detect sabotage?LLM agents and tool use → Agent threats, guardrails and governance
Multilevel and Sequential Monte Carlo for Training-Free Diffusion GuidanceDiffusion models → Diffusion foundations and sampling
Safety Generalization Under Distribution Shift in Safe Reinforcement Learning: A Diabetes TestbedDeep reinforcement learning methods → Safe and constrained RL
Deep Reinforcement Learning for Fault-Adaptive Routing in Eisenstein-Jacobi Interconnection TopologiesDeep reinforcement learning methods → RL for energy, networks and operations
OpenSec: Measuring Incident Response Agent Calibration Under Adversarial EvidenceLLM agents and tool use → Agent threats, guardrails and governance
Shape of Thought: Progressive Object Assembly via Visual Chain-of-ThoughtLLM reasoning → Multimodal reasoning
Towards Mitigating Modality Bias in Vision-Language Models for Temporal Action LocalizationMultimodal LLMs / vision-language models → Video understanding with MLLMs
Llama-3.1-FoundationAI-SecurityLLM-Reasoning-8B Technical ReportLLM reasoning → Reasoning for agents, planning and code
Thinking in Frames: How Visual Context and Test-Time Scaling Empower Video ReasoningLLM reasoning → Multimodal reasoning +1
Predict-Project-Renoise: Sampling Diffusion Models under Hard ConstraintsDiffusion models → Diffusion foundations and sampling
A Diffusive Classification Loss for Learning Energy-based Generative ModelsDiffusion models → Diffusion foundations and sampling
UrduBench: An Urdu Reasoning Benchmark using Contextually Ensembled Translations with Human-in-the-LoopLLM reasoning → Reasoning benchmarks and evaluation
Distributional Active InferenceDeep reinforcement learning methods → Deep RL algorithms and representations
DeepSearchQA: Bridging the Comprehensiveness Gap for Deep Research AgentsLLM agents and tool use → Web, search, and research agents
ICON: Intent-Context Coupling for Efficient Multi-Turn Jailbreak AttackLLM safety and security → Jailbreaking
FreeFix: Boosting 3D Gaussian Splatting via Fine-Tuning-Free Diffusion ModelsDiffusion models → 3D, point cloud and scene generation
SokoBench: Evaluating Long-Horizon Planning and Reasoning in Large Language ModelsLLM reasoning → Reasoning for agents, planning and code
End-to-end example-based sim-to-real RL policy transfer based on neural stylisation with application to robotic cuttingDeep reinforcement learning methods → RL for robot locomotion and manipulation
Reward Models Inherit Value Biases from PretrainingLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
MemCtrl: Using MLLMs as Active Memory Controllers on Embodied AgentsLLM agents and tool use → Agent memory systems
Training Reasoning Models on Saturated Problems via Failure-Prefix ConditioningLLM reasoning → Reinforcement learning for reasoning
Reinforcement Learning via Self-DistillationKnowledge distillation → On-policy and self-distillation for LLMs +1
Dissecting Multimodal In-Context Learning: Modality Asymmetries and Circuit Dynamics in modern TransformersMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
FAIRT2V: Training-Free Debiasing for Text-to-Video Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
SERA: Soft-Verified Efficient Repository AgentsLLM agents and tool use → Software engineering coding agents
REASON: Accelerating Probabilistic Logical Reasoning for Scalable Neuro-Symbolic IntelligenceLLM reasoning → Reasoning benchmarks and evaluation
Less is More: Clustered Cross-Covariance Control for Offline RLOffline RL and imitation learning from data → Offline RL algorithms
Agentic Fog: A Policy-driven Framework for Distributed Intelligence in Fog ComputingLLM agents and tool use → General LLM agent frameworks and benchmarks
Exploring Re-inforcement Learning via Human Feedback under User HeterogeneityLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Persona Prompting as a Lens on LLM Social ReasoningLLM reasoning → Reasoning benchmarks and evaluation +1
Supervised Guidance Training for Infinite-Dimensional Diffusion ModelsDiffusion models → Inverse problems and image restoration
GraphAllocBench: A Flexible Benchmark for Preference-Conditioned Multi-Objective Policy LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
Continual GUI AgentsLLM agents and tool use → GUI and mobile agents
AgentLongBench: A Controllable Long Benchmark For Long-Contexts Agents via Environment RolloutsLLM agents and tool use → Agent memory systems
Adapting the Behavior of Reinforcement Learning Agents to Changing Action Spaces and Reward FunctionsDeep reinforcement learning methods → Deep RL algorithms and representations
MedViz: An Agent-based, Visual-guided Research Assistant for Navigating Biomedical LiteratureLLM agents and tool use → Medical and scientific domain agents
NPU Design for Diffusion Language Model InferenceDiffusion models → Discrete and text diffusion +1
Decoupling Perception and Calibration: Label-Efficient Image Quality Assessment FrameworkMultimodal LLMs / vision-language models → Image editing and quality assessment
Positive-Unlabeled Reinforcement Learning Distillation for On-Premise Small ModelsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
bi-modal textual prompt learning for vision-language models in remote sensingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Harnessing Large Language Models for Precision Querying and Retrieval-Augmented Knowledge Extraction in Clinical Data ScienceRetrieval-augmented generation → Medical and biomedical RAG
GPO: Growing Policy Optimization for Legged Robot Locomotion and Whole-Body ControlDeep reinforcement learning methods → RL for robot locomotion and manipulation
OS-Marathon: Benchmarking Computer-Use Agents on Vast-Horizon, Repetitive TasksLLM agents and tool use → GUI and mobile agents
P2S: Probabilistic Process Supervision for General-Domain Reasoning Question AnsweringLLM reasoning → Reinforcement learning for reasoning
Detecting and Mitigating Memorization in Diffusion Models through Anisotropy of the Log-ProbabilityDiffusion models → Diffusion foundations and sampling
Investigating the Development of Task-Oriented Communication in Vision-Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
When Vision Meets Texts in Listwise RerankingMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Agent Benchmarks Fail Public Sector RequirementsLLM agents and tool use → General LLM agent frameworks and benchmarks
DRAINCODE: Stealthy Energy Consumption Attacks on Retrieval-Augmented Code Generation via Context PoisoningRetrieval-augmented generation → Privacy and security of RAG
Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question ReformulationLLM reasoning → Reinforcement learning for reasoning
AgentIF-OneDay: A Task-level Instruction-Following Benchmark for General AI Agents in Daily ScenariosLLM agents and tool use → General LLM agent frameworks and benchmarks
WFR-MFM: One-Step Inference for Dynamic Unbalanced Optimal TransportDiffusion models → Flow matching, time series and science
Dialogical Reasoning Across AI Architectures: A Multi-Model Framework for Testing AI Alignment StrategiesLLM reasoning → Reasoning benchmarks and evaluation
CLEAR-Mamba:Towards Accurate, Adaptive and Trustworthy Multi-Sequence Ophthalmic Angiography ClassificationState-space sequence models (S4, Mamba) → Mamba for medical image analysis
Ranking-aware Reinforcement Learning for Ordinal RankingDeep reinforcement learning methods → RL training of language and agent models
Reinforcement Unlearning via Group Relative Policy OptimizationLLM safety and security → Machine unlearning
DiffVC-RT: Towards Practical Real-Time Diffusion-based Perceptual Neural Video CompressionDiffusion models → Diffusion for compression and wireless
Advancing Open-source World ModelsWorld models → Video-generation world models and simulators
PathWise: Planning through World Model for Automated Heuristic Design via Self-Evolving LLMsWorld models → World models for language agents +1
Interpreting Emergent Extreme Events in Multi-Agent SystemsLLM agents and tool use → Social simulation and multi-agent debate
Context Tokens are Anchors: Understanding the Repetition Curse in dMLLMs from an Information Flow PerspectiveMultimodal LLMs / vision-language models → Visual token pruning and efficiency
CCMamba: Topologically-Informed Selective State-Space Networks on Combinatorial Complexes for Higher-Order Graph LearningState-space sequence models (S4, Mamba) → Video, motion and generative Mamba models
Efficient Autoregressive Video Diffusion with Dummy HeadDiffusion models → Video generation
Spectral Diffusion Models on the SphereDiffusion models → Diffusion foundations and sampling
CtrlCoT: Dual-Granularity Chain-of-Thought Compression for Controllable ReasoningLLM reasoning → Test-time compute and efficient thinking
BMAM: Brain-inspired Multi-Agent Memory FrameworkLLM agents and tool use → Agent memory systems
PEARL: Plan Exploration and Adaptive Reinforcement Learning for Multihop Tool UseLLM agents and tool use → Reinforcement learning for LLM agents
MARE: Multimodal Alignment and Reinforcement for Explainable Deepfake Detection via Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Mix2Morph: Learning Sound Morphing from Noisy MixesDiffusion models → Audio, speech and music
Concept Component Analysis: A Principled Approach for Concept Extraction in LLMsMechanistic interpretability → SAE features and their applications
Let's Roll a BiFTA: Bi-refinement for Fine-grained Text-visual Alignment in Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Beyond Accuracy: A Cognitive Load Framework for Mapping the Capability Boundaries of Tool-use AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
On the Impact of AGENTS.md Files on the Efficiency of AI Coding AgentsLLM agents and tool use → Software engineering coding agents
Eliminating Hallucination in Diffusion-Augmented Interactive Text-to-Image RetrievalDiffusion models → Image generation, text-to-image and editing
Reinforcement Learning for Dividend Optimization in Partially Observed Regime-Switching Diffusion ModelDeep reinforcement learning methods → Adversarial robustness and RL in finance
OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task ExecutionLLM agents and tool use → GUI and mobile agents +1
Policy of Thoughts: Scaling Test-Time Training for LLM Reasoning via Online Policy EvolutionLLM reasoning → Test-time compute and efficient thinking
LLM-AutoDP: Automatic Data Processing via LLM Agents for Model Fine-tuningLLM agents and tool use → Web, search, and research agents
RAW-Flow: Advancing RGB-to-RAW Image Reconstruction with Deterministic Latent Flow MatchingDiffusion models → Inverse problems and image restoration
Can Continuous-Time Diffusion Models Generate and Solve Globally Constrained Discrete Problems? A Study on SudokuDiffusion models → Discrete and text diffusion
TABED: Test-Time Adaptive Ensemble Drafting for Robust Speculative Decoding in LVLMsEfficient LLM inference and serving → Speculative decoding
AMA: Adaptive Memory via Multi-Agent CollaborationLLM agents and tool use → Agent memory systems
Improving Diffusion Language Model Decoding through Joint Search in Generation Order and Token SpaceDiffusion models → Discrete and text diffusion
MobileBench-OL: A Comprehensive Chinese Benchmark for Evaluating Mobile GUI Agents in Real-World EnvironmentLLM agents and tool use → GUI and mobile agents
Demonstration-Free Robotic Control via LLM AgentsVision-language-action models → Failure recovery and robustness +1
Window-Diffusion: Accelerating Diffusion Language Model Inference with Windowed Token Pruning and CachingDiffusion models → Discrete and text diffusion +1

All 404