Papers

New on arXiv

Thu 12 Feb 2026

446 papers

SPIRE: Structure-Preserving Interpretable Retrieval of EvidenceRetrieval-augmented generation → Multimodal RAG
MoE-SpAc: Efficient MoE Inference Based on Speculative Activation Utility in Heterogeneous Edge ScenariosMixture-of-experts → Efficient MoE training and inference systems
Visual Foresight for Robotic Stow: A Diffusion-Based World Model from Sparse SnapshotsWorld models → Embodied and robot world models
FireRed-Image-Edit-1.0 Technical ReportDiffusion models → Image generation, text-to-image and editing
Continuous Diffusion Models Can Obey Formal SyntaxDiffusion models → Discrete and text diffusion
Safe Reinforcement Learning via Recovery-based Shielding with Gaussian Process Dynamics ModelsDeep reinforcement learning methods → Safe and constrained RL
Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path ForwardLLM agents and tool use → Agent skill learning and evolution
CacheMind: From Miss Rates to Why -- Natural-Language, Trace-Grounded Reasoning for Cache ReplacementLLM reasoning → Test-time compute and efficient thinking +1
Sparse Autoencoders are Capable LLM Jailbreak MitigatorsLLM safety and security → Jailbreaking +1
Self-Refining Vision Language Model for Robotic Failure Detection and ReasoningLLM reasoning → Spatial and embodied reasoning +1
MonoLoss: A Training Objective for Interpretable Monosemantic RepresentationsMechanistic interpretability → SAE features and their applications
AstRL: Analog and Mixed-Signal Circuit Synthesis with Deep Reinforcement LearningDeep reinforcement learning methods → RL for energy, networks and operations
What does RL improve for Visual Reasoning? A Frankenstein-Style AnalysisLLM reasoning → Multimodal reasoning +1
Reproducing DragDiffusion: Interactive Point-Based Editing with Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
Value Bonuses using Ensemble Errors for Exploration in Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Wireless TokenCom: RL-Based Tokenizer Agreement for Multi-User Wireless Token CommunicationsDeep reinforcement learning methods → RL for energy, networks and operations
ForeAct: Steering Your VLA with Efficient Visual Foresight PlanningVision-language-action models → World model-based VLA
Abstractive Red-Teaming of Language Model CharacterLLM safety and security → Prompt injection and agent security
AgenticShop: Benchmarking Agentic Product Curation for Personalized Web ShoppingLLM agents and tool use → Social simulation and multi-agent debate
Perceptual Self-Reflection in Agentic Physics Simulation Code GenerationLLM agents and tool use → Multimodal and embodied agents
Scaling Verification Can Be More Effective than Scaling Policy Learning for Vision-Language-Action AlignmentVision-language-action models → Failure recovery and robustness
UniT: Unified Multimodal Chain-of-Thought Test-time ScalingLLM reasoning → Multimodal reasoning
AttentionRetriever: Attention Layers are Secretly Long Document RetrieversRetrieval-augmented generation → Generation and context integration
Agentic Test-Time Scaling for WebAgentsLLM reasoning → Reasoning for agents, planning and code +1
On-Policy Context Distillation for Language ModelsKnowledge distillation → On-policy and self-distillation for LLMs
Function-Space Decoupled Diffusion for Forward and Inverse Modeling in Carbon Capture and StorageDiffusion models → Flow matching, time series and science
MonarchRT: Efficient Attention for Real-Time Video GenerationDiffusion models → Video generation
CM2: Reinforcement Learning with Checklist Rewards for Multi-Turn and Multi-Step Agentic Tool UseLLM agents and tool use → Reinforcement learning for LLM agents
Few-Step Diffusion Language Models via Trajectory Self-DistillationDiffusion models → Discrete and text diffusion +2
Think like a Scientist: Physics-guided LLM Agent for Equation DiscoveryLLM reasoning → Knowledge- and tool-augmented reasoning +1
Detecting Overflow in Compressed Token Representations for Retrieval-Augmented GenerationRetrieval-augmented generation → Generation and context integration
Categorical Flow MapsDiffusion models → Flow matching, time series and science +1
Diffusion Alignment Beyond KL: Variance Minimisation as Effective Policy OptimiserDiffusion models → Diffusion and flow policies for robotics
Best of Both Worlds: Multimodal Reasoning and Generation via Unified Discrete Flow MatchingLLM reasoning → Multimodal reasoning +1
The Observer Effect in World Models: Invasive Adaptation Corrupts Latent PhysicsWorld models → Latent predictive world models
Visual Reasoning Benchmark: Evaluating Multimodal LLMs on Classroom-Authentic Visual Problems from Primary EducationLLM reasoning → Multimodal reasoning +1
MalTool: Malicious Tool Attacks on LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
Convex Markov Games and Beyond: New Proof of Existence, Characterization and Learning Algorithms for Nash EquilibriaDeep reinforcement learning methods → RL for games and combinatorial optimization
How Sampling Shapes LLM Alignment: From One-Shot Optima to Iterative DynamicsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Pedagogically-Inspired Data Synthesis for Language Model Knowledge DistillationSynthetic data and self-improvement for LLMs → Code and instruction data synthesis +1
Sci-CoE: Co-evolving Scientific Reasoning LLMs via Geometric Consensus with Sparse SupervisionLLM reasoning → Mathematical reasoning
3DGSNav: Enhancing Vision-Language Model Reasoning for Object Navigation via Active 3D Gaussian SplattingLLM reasoning → Spatial and embodied reasoning +1
SafeNeuron: Neuron-Level Safety Alignment for Large Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +2
TexSpot: 3D Texture Enhancement with Spatially-uniform Point Latent RepresentationDiffusion models → 3D, point cloud and scene generation
FAIL: Flow Matching Adversarial Imitation Learning for Image GenerationDiffusion models → Diffusion and flow policies for robotics +2
OServe: Accelerating LLM Serving via Spatial-Temporal Workload OrchestrationEfficient LLM inference and serving → Serving systems and request scheduling
Seq2Seq2Seq: Lossless Data Compression via Discrete Latent Transformers and Reinforcement LearningDeep reinforcement learning methods → RL training of language and agent models
On the Adoption of AI Coding Agents in Open-source Android and iOS DevelopmentLLM agents and tool use → Software engineering coding agents
STAR : Bridging Statistical and Agentic Reasoning for Large Model Performance PredictionLLM reasoning → Reasoning for agents, planning and code +1
BlackCATT: Black-box Collusion Aware Traitor Tracing in Federated LearningLLM safety and security → Federated unlearning and poisoning
WavBench: Benchmarking Reasoning, Colloquialism, and Paralinguistics for End-to-End Spoken Dialogue ModelsLLM reasoning → Audio-language model reasoning
Value Alignment Tax: Measuring Value Trade-offs in LLM AlignmentLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Learning beyond Teacher: Generalized On-Policy Distillation with Reward ExtrapolationKnowledge distillation → On-policy and self-distillation for LLMs +1
Alignment Risks from Capability-Seeking RL TrainingLLM safety and security → Safety alignment, refusal and guardrails
P-GenRM: Personalized Generative Reward Model with Test-time User-based ScalingLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Stop Unnecessary Reflection: Training LRMs for Efficient Reasoning with Adaptive Reflection and Length Coordinated PenaltyLLM reasoning → Test-time compute and efficient thinking
On the Complexity of Offline Reinforcement Learning with Q^star-Approximation and Partial CoverageOffline RL and imitation learning from data → Offline RL algorithms
GigaBrain-0.5M*: a VLA That Learns From World Model-Based Reinforcement LearningWorld models → Embodied and robot world models +1
Choose Your Agent: Tradeoffs in Adopting AI Advisors, Coaches, and Delegates in Multi-Party NegotiationLLM agents and tool use → Social simulation and multi-agent debate
Geometry of Uncertainty: Learning Metric Spaces for Multimodal State Estimation in RLDeep reinforcement learning methods → Deep RL algorithms and representations
Differentiable Modal Logic for Multi-Agent Diagnosis, Orchestration and CommunicationLLM agents and tool use → General LLM agent frameworks and benchmarks
Tiny Recursive Reasoning with Mamba-2 Attention HybridLLM reasoning → Test-time compute and efficient thinking +1
VLAW: Iterative Co-Improvement of Vision-Language-Action Policy and World ModelWorld models → Embodied and robot world models +1
HoloBrain-0 Technical ReportVision-language-action models → Generalist VLA models and benchmarks
LawThinker: A Deep Research Legal Agent in Dynamic EnvironmentsLLM reasoning → Reasoning for agents, planning and code
Fourier Transformers for Latent Crystallographic Diffusion and Generative ModelingDiffusion models → Flow matching, time series and science
PrefillShare: A Shared Prefill Module for KV Reuse in Multi-LLM Disaggregated ServingEfficient LLM inference and serving → Serving systems and request scheduling
Protein Circuit Tracing via Cross-layer TranscodersMechanistic interpretability → Interpreting biomedical and science models
InjectRBP: Steering Large Language Model Reasoning Behavior via Pattern InjectionLLM reasoning → Test-time compute and efficient thinking
Can Local Vision-Language Models improve Activity Recognition over Vision Transformers? -- Case Study on Newborn ResuscitationMultimodal LLMs / vision-language models → Medical vision-language models
Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?LLM agents and tool use → Software engineering coding agents
Spatial Chain-of-Thought: Bridging Understanding and Generation Models for Spatial Reasoning GenerationLLM reasoning → Multimodal reasoning +1
Accelerating Robotic Reinforcement Learning with Agent GuidanceDeep reinforcement learning methods → RL for robot locomotion and manipulation
Gaia2: Benchmarking LLM Agents on Dynamic and Asynchronous EnvironmentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Wisdom of the LLM Crowd: A Large Scale Benchmark of Multi-Label U.S. Election-Related Harmful Social Media ContentLLM safety and security → Safety alignment, refusal and guardrails
Benchmarking Vision-Language Models for French PDF-to-Markdown ConversionMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Synthesis of Late Gadolinium Enhancement Images via Implicit Neural Representations for Cardiac Scar SegmentationDiffusion models → Medical imaging and segmentation
Extending Puzzle for Mixture-of-Experts Reasoning Models with Application to GPT-OSS AccelerationLLM reasoning → Test-time compute and efficient thinking +1
AdaptEvolve: Improving Efficiency of Evolutionary AI Agents through Adaptive Model SelectionLLM reasoning → Reasoning for agents, planning and code +1
TADA! Tuning Audio Diffusion Models through Activation SteeringDiffusion models → Audio, speech and music +1
Echo: Towards Advanced Audio Comprehension via Audio-Interleaved ReasoningLLM reasoning → Audio-language model reasoning
Mitigating Mismatch within Reference-based Preference OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Agentic AI for Cybersecurity: A Meta-Cognitive Architecture for Governable AutonomyLLM agents and tool use → Agent threats, guardrails and governance
Where Bits Matter in World Model Planning: A Paired Mixed-Bit Study for Efficient Spatial ReasoningWorld models → World models for language agents +1
From Atoms to Trees: Building a Structured Feature Forest with Hierarchical Sparse AutoencodersMechanistic interpretability → SAE features and their applications
DiffPlace: Street View Generation via Place-Controllable Diffusion Model Enhancing Place RecognitionDiffusion models → Diffusion and flow policies for robotics
LAMP: Implicit Language Map for Robot NavigationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Talk2DM: Enabling Natural Language Querying and Commonsense Reasoning for Vehicle-Road-Cloud Integrated Dynamic Maps with Large Language ModelsLLM reasoning → Chain-of-thought reasoning for driving
Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal PerceptionMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Free Lunch for Stabilizing Rectified Flow InversionDiffusion models → Image generation, text-to-image and editing
JEPA-VLA: Video Predictive Embedding is Needed for VLA ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Revis: Sparse Latent Steering to Mitigate Object Hallucination in Large Vision-Language ModelsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Predicting LLM Output Length via Entropy-Guided RepresentationsEfficient LLM inference and serving → Serving systems and request scheduling
Deep Kernel Fusion for TransformersLLM agents and tool use → Software engineering coding agents
PuYun-LDM: A Latent Diffusion Model for High-Resolution Ensemble Weather ForecastsDiffusion models → Flow matching, time series and science
From Path Signatures to Sequential Modeling: Incremental Signature Contributions for Offline RLOffline RL and imitation learning from data → Decision transformers and sequence modeling
Temporal Difference Learning with Constrained Initial RepresentationsDeep reinforcement learning methods → Deep RL algorithms and representations
More Haste, Less Speed: Weaker Single-Layer Watermark Improves Distortion-Free Watermark EnsemblesLLM safety and security → Watermarking generated content and models
Detecting RLVR Training Data via Structural Convergence of ReasoningLLM reasoning → Reinforcement learning for reasoning
Beyond End-to-End Video Models: An LLM-Based Multi-Agent System for Educational Video GenerationLLM agents and tool use → Multimodal and embodied agents

All 446