Papers

New on arXiv

Sat 31 Jan 2026

316 papers

Exploring Structural Complexity in Normative RAG with Graph-based approaches: A case study on the ETSI StandardsRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
MCERF: Advancing Multimodal LLM Evaluation of Engineering Documentation with Enhanced RetrievalRetrieval-augmented generation → Multimodal RAG
RLHF May Not Reflect Genuine PreferencesLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Supporting Calibrated Reliance in Human-AI Collaboration: Different Strategies for Different TasksLLM reasoning → Reasoning benchmarks and evaluation
CoEditor++: Instruction-based Visual Editing via Cognitive ReasoningMultimodal LLMs / vision-language models → Image editing and quality assessment
Beyond Static Instruction: A Multi-agent AI Framework for Adaptive Augmented Reality Robot TrainingLLM agents and tool use → Multimodal and embodied agents
Reasoning-Based Personalized Generation for Users with Sparse DataLLM reasoning → Reinforcement learning for reasoning
EPSVec: Efficient and Private Synthetic Data Generation via Dataset VectorsSynthetic data and self-improvement for LLMs → Privacy-preserving tabular data synthesis
Test-Time Adaptation for Tactile-Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
GRACE: an Agentic AI for Particle Physics Experiment Design and SimulationLLM agents and tool use → Web, search, and research agents
Steering to Say No: Configurable Refusal via Activation Steering in Vision Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +2
A Causal Perspective for Enhancing Jailbreak Attack and DefenseLLM safety and security → Jailbreaking
Agentic Observability: Automated Alert Triage for Adobe E-CommerceLLM agents and tool use → Agent threats, guardrails and governance
QuantLRM: Quantization of Large Reasoning Models via Fine-Tuning SignalsLLM reasoning → Test-time compute and efficient thinking
ProphetKV: User-Query-Driven Selective Recomputation for Efficient KV Cache Reuse in Retrieval-Augmented GenerationRetrieval-augmented generation → Generation and context integration +1
Reward Shaping for (Inference-Time) Alignment: A Stackelberg Game PerspectiveLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Improving Minimax Estimation Rates for Contaminated Mixture of Multinomial Logistic Experts via Expert HeterogeneityMixture-of-experts → MoE theory and gating functions
MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP ServersLLM agents and tool use → Tool learning and function calling
Continuous-Utility Direct Preference OptimizationLLM reasoning → Mathematical reasoning +1
Learning Abstractions for Hierarchical Planning in Program-Synthesis AgentsLLM agents and tool use → Reinforcement learning for LLM agents
Beyond What Seems Necessary: Hidden Gains from Scaling Training-Time Reasoning Length under Outcome SupervisionLLM reasoning → Test-time compute and efficient thinking
Supervised sparse auto-encoders for interpretable and compositional representationsMechanistic interpretability → Sparse autoencoder training and evaluation
Green-VLA: Staged Vision-Language-Action Model for Generalist RobotsVision-language-action models → Generalist VLA models and benchmarks
UniMorphGrasp: Diffusion Model with Morphology-Awareness for Cross-Embodiment Dexterous Grasp GenerationDiffusion models → Diffusion and flow policies for robotics
EffGen: Enabling Small Language Models as Capable Autonomous AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
RoDiF: Robust Direct Fine-Tuning of Diffusion Policies with Corrupted Human FeedbackDiffusion models → Diffusion and flow policies for robotics +1
Reliability-Aware Determinantal Point Processes for Robust Informative Data Selection in Large Language ModelsEfficient LLM inference and serving → Early exit and edge inference
DIAMOND: Directed Inference for Artifact Mitigation in Flow Matching ModelsDiffusion models → Flow matching, time series and science
Dynamic Expert Sharing: Decoupling Memory from Parallelism in Mixture-of-Experts Diffusion LLMsMixture-of-experts → Efficient MoE training and inference systems +1
Improving Flow Matching by Aligning Flow DivergenceDiffusion models → Flow matching, time series and science
Unifying Adversarial Robustness and Training Across Text Scoring ModelsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
Investigating the Robustness of Subtask Distillation under Spurious CorrelationKnowledge distillation → General knowledge distillation methods
Omni-RRM: Advancing Omni Reward Modeling via Automatic Rubric-Grounded Preference SynthesisMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
Optimizing Agentic Reasoning with Retrieval via Synthetic Semantic Information Gain RewardLLM reasoning → Knowledge- and tool-augmented reasoning +1
Safety-Efficacy Trade Off: Robustness against Data-PoisoningLLM safety and security → Backdoor and data poisoning attacks
Resource-Efficient Reinforcement for Reasoning Large Language Models via Dynamic One-Shot Policy RefinementLLM reasoning → Reinforcement learning for reasoning
Generating a Paracosm for Training-Free Zero-Shot Composed Image RetrievalMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Physics-informed Diffusion Mamba Transformer for Real-world DrivingDiffusion models → Diffusion and flow policies for robotics +2
Any3D-VLA: Enhancing VLA Robustness via Diverse Point CloudsVision-language-action models → 3D and spatial-aware VLA
JTok: On Token Embedding as another Axis of Scaling Law via Joint Token Self-modulationMixture-of-experts → Expert routing strategies
Latent Shadows: The Gaussian-Discrete Duality in Masked DiffusionDiffusion models → Discrete and text diffusion
World Models as an Intermediary between Agents and the Real WorldWorld models → World models for language agents +1
Fast Non-Episodic Finite-Horizon RL with K-Step Lookahead ThresholdingDeep reinforcement learning methods → RL theory and sample complexity
Environment-Aware Adaptive Pruning with Interleaved Inference Orchestration for Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation
Reasoning as State Transition: A Representational Analysis of Reasoning Evolution in Large Language ModelsLLM reasoning → Reasoning benchmarks and evaluation
Communications-Incentivized Collaborative Reasoning in NetGPT through Agentic Reinforcement LearningLLM agents and tool use → Agentic AI for wireless networks
WordCraft: Scaffolding the Keyword Method for L2 Vocabulary Learning with Multimodal LLMsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
APR: Penalizing Structural Redundancy in Large Reasoning Models via Anchor-based Process RewardsLLM reasoning → Test-time compute and efficient thinking
Adaptive Ability Decomposing for Unlocking Large Reasoning Model Effective Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning
Evolving Interpretable Constitutions for Multi-Agent CoordinationLLM agents and tool use → Social simulation and multi-agent debate +1
Bypassing Prompt Injection Detectors through Evasive InjectionsLLM safety and security → Prompt injection and agent security
HyperOffload: Graph-Driven Hierarchical Memory Management for Large Language Models on SuperNode ArchitecturesLLM reasoning → Test-time compute and efficient thinking
Can Vision-Language Models Handle Long-Context Code? An Empirical Study on Visual CompressionMultimodal LLMs / vision-language models → Visual token pruning and efficiency
SA-VLA: Spatially-Aware Flow-Matching for Vision-Language-Action Reinforcement LearningVision-language-action models → Efficient VLA inference and adaptation +1
MedTextWeaver: Procedural Knowledge Evolution in Agentic Medical Text EditingLLM agents and tool use → Medical and scientific domain agents
Pareto-Conditioned Diffusion Models for Offline Multi-Objective OptimizationDiffusion models → Diffusion foundations and sampling
Supervised makeup transfer with a curated dataset: Decoupling identity and makeup features for enhanced transformationDiffusion models → Image generation, text-to-image and editing
Rethinking Hallucinations: Correctness, Consistency, and Prompt MultiplicityLLM reasoning → Hallucination detection and mitigation
Emergence of Distortions in High-Dimensional Guided Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Self-Guard: Defending Large Reasoning Models via enhanced self-reflectionLLM reasoning → Reasoning for agents, planning and code
Learning to Accelerate Vision-Language-Action Models through Adaptive Visual Token CachingVision-language-action models → Efficient VLA inference and adaptation
HumanStudy-Bench: Towards AI Agent Design for Participant SimulationLLM agents and tool use → Social simulation and multi-agent debate
Video Understanding: Through A Temporal LensMultimodal LLMs / vision-language models → Video understanding with MLLMs
Toward Reliable Sim-to-Real Predictability for MoE-based Robust Quadrupedal LocomotionDeep reinforcement learning methods → RL for robot locomotion and manipulation
Can Small Language Models Handle Context-Summarized Multi-Turn Customer-Service QA? A Synthetic Data-Driven Comparative EvaluationSynthetic data and self-improvement for LLMs → Synthetic reasoning data and bootstrapping
SEISMO: Explanation-Aware, Trajectory-Conditioned LLM Agents for Sample-Efficient Molecular OptimisationLLM agents and tool use → Medical and scientific domain agents
DisRFM: Polar Riemannian Flow Matching for Structure-Preserving Graph Domain AdaptationDiffusion models → Flow matching, time series and science
Non-Contrastive Vision-Language Learning with Predictive Embedding AlignmentMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Solving Room Impulse Response Inverse Problems Using Flow Matching with Analytic Wiener DenoiserDiffusion models → Flow matching, time series and science
A Hybrid Mamba-SAM Architecture for Efficient 3D Medical Image SegmentationState-space sequence models (S4, Mamba) → Mamba for medical image analysis
LegalOne: A Family of Foundation Models for Reliable Legal ReasoningLLM reasoning → Reasoning for agents, planning and code
On the Equilibrium between Feasible Zone and Uncertain Model in Safe ExplorationDeep reinforcement learning methods → Safe and constrained RL
Towards Sample-Efficient and Stable Reinforcement Learning for LLM-based RecommendationLLM reasoning → LLM reasoning for recommendation
Model-Based Data-Efficient and Robust Reinforcement LearningDeep reinforcement learning methods → RL for energy, networks and operations
Action-Free Offline-to-Online RL via Discretised State PoliciesOffline RL and imitation learning from data → Offline RL algorithms
FaceSnap: Enhanced ID-fidelity Network for Tuning-free Portrait CustomizationDiffusion models → Image generation, text-to-image and editing
Towards Interpretable Hallucination Analysis and Mitigation in LVLMs via Contrastive Neuron SteeringMechanistic interpretability → Hallucination detection from model internals
Jailbreaking LLMs via CalibrationLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
SPOT: Selective Prompt Projection via Total Variation for Inference-Only Safe Text-to-Image GenerationDiffusion models → Robustness, privacy and adversarial aspects
Structured Self-Consistency:A Multi-Task Evaluation of LLMs on VirtualHomeLLM reasoning → Spatial and embodied reasoning
Scalable Generative Game Engine: Breaking the Resolution Wall via Hardware-Algorithm Co-DesignWorld models → Video-generation world models and simulators
Direct Preference Optimization with Rating Information: Practical Algorithms and Provable GainsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Pix2Fact: When Vision Is Not Enough -- Benchmarking Fine-Grained VQA with Web Verification on High-Resolution Real-World ScenesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
DockSmith: Scaling Reliable Coding Environments via an Agentic Docker BuilderLLM agents and tool use → Software engineering coding agents
SEER: Transformer-based Robust Time Series Forecasting via Automated Patch Enhancement and ReplacementMixture-of-experts → MoE for time series and graphs
Safe Langevin Soft Actor CriticDeep reinforcement learning methods → RL theory and sample complexity
RAG-GNN: Integrating Retrieved Knowledge with Graph Neural Networks for Precision MedicineRetrieval-augmented generation → Medical and biomedical RAG
SAU: Sparsity-Aware Unlearning for LLMs via Gradient Masking and Importance RedistributionLLM safety and security → Machine unlearning
Agentic Reward Modeling: Verifying GUI Agent via Progressive Trajectory-Grounded InteractionLLM agents and tool use → GUI and mobile agents
Learning Modal-Mixed Chain-of-Thought Reasoning with Latent EmbeddingsLLM reasoning → Multimodal reasoning
Dual-View Predictive Diffusion: Lightweight Speech Enhancement via Spectrogram-Image SynergyDiffusion models → Audio, speech and music
Forget by Uncertainty: Orthogonal Entropy Unlearning for Quantized Neural NetworksLLM safety and security → Machine unlearning
Unmasking Reasoning Processes: A Process-aware Benchmark for Evaluating Structural Mathematical Reasoning in LLMsLLM reasoning → Mathematical reasoning
Learning to Decode Against Compositional Hallucination in Video Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
ConLA: Contrastive Latent Action Learning from Human Videos for Robotic ManipulationVision-language-action models → Human-video and latent-action pretraining
Reasoning by Commented Code for Table Question AnsweringLLM reasoning → Knowledge- and tool-augmented reasoning
Surrogate Ensemble in Expensive Multi-Objective Optimization via Deep Q-LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
How Far Are LLMs from Professional Poker Players? Revisiting Game-Theoretic Reasoning with Agentic Tool UseLLM reasoning → Reasoning for agents, planning and code +1
SAGE: Accelerating Vision-Language Models via Entropy-Guided Adaptive Speculative DecodingMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
MRAD: Zero-Shot Anomaly Detection with Memory-Driven RetrievalMultimodal LLMs / vision-language models → Chart understanding and anomaly detection

All 316