Papers

New on arXiv

Fri 6 Feb 2026

441 papers

Software Self-Extension with SelfEvolve: an Agentic Architecture for Runtime Code GenerationLLM agents and tool use → Software engineering coding agents
VenusBench-Mobile: A Challenging and User-Centric Benchmark for Mobile GUI Agents with Capability DiagnosticsLLM agents and tool use → GUI and mobile agents
UnWeaving the knots of GraphRAG -- turns out VectorRAG is almost enoughRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
TTSR: Test-Time Self-Evolving via ReflectionSynthetic data and self-improvement for LLMs → Synthetic reasoning data and bootstrapping
PlugMem: A Task-Agnostic Plugin Memory Module for LLM AgentsLLM agents and tool use → Agent memory systems
Language Model Goal Selection Differs from Humans' in a Self-Directed Learning TaskLLM agents and tool use → General LLM agent frameworks and benchmarks
SE-Search: Self-Evolving Search Agent via Memory and Dense RewardRetrieval-augmented generation → Agentic RAG, agents and memory +2
From Conflict to Consensus: Boosting Medical Reasoning via Multi-Round Agentic RAGLLM reasoning → Clinical and biomedical applications +1
One Bias After Another: Mechanistic Reward Shaping and Persistent Biases in Language Reward ModelsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
How Well Can LLM Agents Simulate End-User Security and Privacy Attitudes and Behaviors?LLM agents and tool use → Agent threats, guardrails and governance
Toward Self-Driving Universities: Can Universities Drive Themselves with Agentic AI?LLM agents and tool use → Educational agents and math problem solving
The Doctor Will (Still) See You Now: On the Structural Limits of Agentic AI in HealthcareLLM agents and tool use → Medical and scientific domain agents
Can LLM Safety Be Ensured by Constraining Parameter Regions?LLM safety and security → Safety alignment, refusal and guardrails
A Case Study of Selected PTQ Baselines for Reasoning LLMs on Ascend NPULLM reasoning → Test-time compute and efficient thinking
Agentic Unlearning: When LLM Agent Meets Machine UnlearningLLM agents and tool use → Agent memory systems +1
Tethered Reasoning: Decoupling Entropy from Hallucination in Quantized LLMs via Manifold SteeringLLM reasoning → Hallucination detection and mitigation +1
Robust Pre-Training of Medical Vision-and-Language Models with Domain-Invariant Multi-Modal Masked ReconstructionMultimodal LLMs / vision-language models → Medical vision-language models
BEAGLE: Behavior-Enforced Agent for Grounded Learner EmulationLLM agents and tool use → Educational agents and math problem solving
Towards Understanding Multimodal Fine-Tuning: Spatial FeaturesMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
TermiGen: High-Fidelity Environment and Robust Trajectory Synthesis for Terminal AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
XShare: Collaborative in-Batch Expert Sharing for Faster MoE InferenceMixture-of-experts → Efficient MoE training and inference systems +1
Beyond Crash: Hijacking Your Autonomous Vehicle for Fun and ProfitEnd-to-end autonomous driving → Adversarial scenarios and safety testing
ArcMark: Distortion-Free Multi-Byte LLM Watermark via Optimal TransportLLM safety and security → Watermarking generated content and models
Vegas: Self-Speculative Decoding with Verification-Guided Sparse AttentionEfficient LLM inference and serving → Speculative decoding
Beyond Solving: Prescriptive Probing for Neural Routing SolversDeep reinforcement learning methods → RL for games and combinatorial optimization
Multi-Agentic AI for Fairness-Aware and Accelerated Multi-modal Large Model Inference in Real-world Mobile Edge NetworksLLM agents and tool use → General LLM agent frameworks and benchmarks
Adaptive Retrieval helps Reasoning in LLMs -- but mostly if it's not usedLLM reasoning → Knowledge- and tool-augmented reasoning
Risk-Sensitive Exponential Actor CriticDeep reinforcement learning methods → RL theory and sample complexity
Latent Target Score Matching, with an application to Simulation-Based InferenceDiffusion models → Diffusion foundations and sampling
PreFlect: From Retrospective to Prospective Reflection in Large Language Model AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
The Value of Variance: Mitigating Debate Collapse in Multi-Agent Systems via Uncertainty-Driven Policy OptimizationLLM reasoning → Reasoning for agents, planning and code +1
PACIFIC: Can LLMs Discern the Psychometric Traits Influencing Your Preferences? Personality-Driven Preference Alignment in LLMsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Learning Nonlinear Systems In-Context: From Synthetic Data to Real-World Motor ControlSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
ANCHOR: Branch-Point Data Generation for GUI AgentsLLM agents and tool use → GUI and mobile agents
On Randomness in Agentic EvalsLLM agents and tool use → General LLM agent frameworks and benchmarks
Finding Connections: Membership Inference Attacks for the Multi-Table Synthetic Data SettingLLM safety and security → Membership inference attacks
Reasoning-Augmented Representations for Multimodal RetrievalLLM reasoning → Multimodal reasoning
ShallowJail: Steering Jailbreaks against Large Language ModelsLLM safety and security → Jailbreaking
Extended to Reality: Prompt Injection in 3D EnvironmentsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
scDFM: Distributional Flow Matching Model for Robust Single-Cell Perturbation PredictionDiffusion models → Flow matching, time series and science
Fast and Robust Likelihood-Guided Diffusion Posterior Sampling with Amortized Variational InferenceDiffusion models → Inverse problems and image restoration
Lemon Agent Technical ReportLLM agents and tool use → General LLM agent frameworks and benchmarks
MosaicThinker: On-Device Visual Spatial Reasoning for Embodied AI via Iterative Construction of Space RepresentationLLM reasoning → Spatial and embodied reasoning
LatentChem: From Textual CoT to Latent Thinking in Chemical ReasoningLLM reasoning → Clinical and biomedical applications
MedMO: Grounding and Understanding Multimodal Large Language Model for Medical ImagesMultimodal LLMs / vision-language models → Medical vision-language models
Learning a Generative Meta-Model of LLM ActivationsMechanistic interpretability → Interpreting and steering diffusion models
InftyThink+: Effective and Efficient Infinite-Horizon Reasoning via Reinforcement LearningLLM reasoning → Test-time compute and efficient thinking
DAWN: Dependency-Aware Fast Inference for Diffusion LLMsEfficient LLM inference and serving → Diffusion LLM and MoE inference
DreamDojo: A Generalist Robot World Model from Large-Scale Human VideosWorld models → Embodied and robot world models
Agentic Uncertainty Reveals Agentic OverconfidenceLLM agents and tool use → General LLM agent frameworks and benchmarks
Endogenous Resistance to Activation Steering in Language ModelsMechanistic interpretability → Activation steering methods
When RL Meets Adaptive Speculative Training: A Unified Training-Serving SystemEfficient LLM inference and serving → Speculative decoding
Continuous-time reinforcement learning: ellipticity enables model-free value function approximationDeep reinforcement learning methods → RL theory and sample complexity
From Kepler to Newton: Inductive Biases Guide Learned World Models in TransformersWorld models → Latent predictive world models
TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and TamperingLLM safety and security → Jailbreaking
Plato's Form: Toward Backdoor Defense-as-a-Service for LLMs with Prototype RepresentationsLLM safety and security → Backdoor and data poisoning attacks
Prompt Reinjection: Alleviating Prompt Forgetting in Multimodal Diffusion Transformers for Text-to-Image GenerationDiffusion models → Image generation, text-to-image and editing
NanoFLUX: Distillation-Driven Compression of Large Text-to-Image Generation Models for Mobile DevicesDiffusion models → Image generation, text-to-image and editing
TraceCoder: A Trace-Driven Multi-Agent Framework for Automated Debugging of LLM-Generated CodeLLM agents and tool use → Software engineering coding agents
RFDM: Residual Flow Diffusion Model for Efficient Causal Video EditingDiffusion models → Video generation
Uncovering Cross-Objective Interference in Multi-Objective AlignmentLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Parameters as Experts: Adapting Vision Models with Dynamic Parameter RoutingMixture-of-experts → Vision mixture-of-experts
AIRS-Bench: a Suite of Tasks for Frontier AI Research Science AgentsLLM agents and tool use → Web, search, and research agents
SEMA: Simple yet Effective Learning for Multi-Turn Jailbreak AttacksLLM safety and security → Jailbreaking
The Quantum Sieve Tracer: A Hybrid Framework for Layer-Wise Activation Tracing in Large Language ModelsMechanistic interpretability → Circuits and mechanistic analysis
Improved Sampling Schedules for Discrete Diffusion ModelsDiffusion models → Diffusion foundations and sampling
DynFOA: Generating First-Order Ambisonics with Conditional Diffusion for Dynamic and Acoustically Complex 360-Degree VideosDiffusion models → Audio, speech and music
From Features to Actions: Explainability in Traditional and Agentic AI SystemsLLM agents and tool use → Software engineering coding agents
LLM Active Alignment: A Nash Equilibrium PerspectiveLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
AEGPO: Adaptive Entropy-Guided Policy Optimization for Diffusion ModelsDiffusion models → Diffusion and flow policies for robotics +1
RAIGen: Rare Attribute Identification in Text-to-Image Generative ModelsDiffusion models → Image generation, text-to-image and editing
On the Non-Identifiability of Steering Vectors in Large Language ModelsMechanistic interpretability → Activation steering methods
FlowDA: Accurate, Low-Latency Weather Data Assimilation via Flow MatchingDiffusion models → Flow matching, time series and science
Generating Data-Driven Reasoning Rubrics for Domain-Adaptive Reward ModelingLLM reasoning → Reinforcement learning for reasoning +1
Displacement-Resistant Extensions of DPO with Nonconvex f-DivergencesLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
AEGIS: Adversarial Target-Guided Retention-Data-Free Robust Concept Erasure from Diffusion ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
Soft Forward-Backward Representations for Zero-shot Reinforcement Learning with General UtilitiesDeep reinforcement learning methods → Deep RL algorithms and representations
R-Align: Enhancing Generative Reward Models through Rationale-Centric Meta-JudgingLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
A Unified Framework for LLM WatermarksLLM safety and security → Watermarking generated content and models
Beyond Function-Level Analysis: Context-Aware Reasoning for Inter-Procedural Vulnerability DetectionLLM reasoning → Reasoning for agents, planning and code
Semantically Labelled Automata for Multi-Task Reinforcement Learning with LTL InstructionsDeep reinforcement learning methods → Interpretable and logic-guided RL
Table-as-Search: Formulate Long-Horizon Agentic Information Seeking as Table CompletionLLM agents and tool use → Web, search, and research agents
F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the RareLLM reasoning → Reinforcement learning for reasoning
PrefIx: Understand and Adapt to User Preference in Human-Agent InteractionLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Crowd-FM: Learned Optimal Selection of Conditional Flow Matching-generated Trajectories for Crowd NavigationDiffusion models → Diffusion and flow policies for robotics
Memory-Conditioned Flow-Matching for Stable Autoregressive PDE RolloutsDiffusion models → Flow matching, time series and science
compar:IA: The French Government's LLM arena to collect French-language human prompts and preference dataLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
PlanViz: Evaluating Planning-Oriented Image Generation and Editing for Computer-Use TasksLLM reasoning → Multimodal reasoning
Same Answer, Different Representations: Hidden instability in VLMsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Beyond Static Alignment: Hierarchical Policy Control for LLM Safety via Risk-Aware Chain-of-ThoughtLLM reasoning → Reasoning for agents, planning and code +2
TrapSuffix: Proactive Defense Against Adversarial Suffixes in JailbreakingLLM safety and security → Jailbreaking
Do Prompts Guarantee Safety? Mitigating Toxicity from LLM Generations through Subspace InterventionLLM safety and security → Safety alignment, refusal and guardrails
Infinite-dimensional generative diffusions via Doob's h-transformDiffusion models → Diffusion foundations and sampling
Force Generative Imitation Learning: Bridging Position Trajectory and Force Commands through Control TechniqueOffline RL and imitation learning from data → Robot manipulation imitation learning
CauCLIP: Bridging the Sim-to-Real Gap in Surgical Video Understanding via Causality-Inspired Vision-Language ModelingMultimodal LLMs / vision-language models → Medical vision-language models
Confundo: Learning to Generate Robust Poison for Practical RAG SystemsRetrieval-augmented generation → Privacy and security of RAG
The hidden risks of temporal resampling in clinical reinforcement learningOffline RL and imitation learning from data → Safe offline RL +1
Echoes as Anchors: Probabilistic Costs and Attention Refocusing in LLM ReasoningLLM reasoning → Test-time compute and efficient thinking
Sample-Efficient Policy Space Response Oracles with Joint Experience Best ResponseDeep reinforcement learning methods → RL theory and sample complexity
DiTS: Multimodal Diffusion Transformers Are Time Series ForecastersDiffusion models → Flow matching, time series and science

All 441