Papers

New on arXiv

Thu 4 Dec 2025

351 papers

KV Cache Recycling to Expand Usable Context Capacity in Low Parameter LLMsEfficient LLM inference and serving → KV cache optimization
Tiny Recursive Models on ARC-AGI-1: Inductive Biases, Identity Conditioning, and Test-Time ComputeLLM reasoning → Multimodal reasoning
Norm-Governed Multi-Agent Decision-Making in Simulator-Coupled Environments:The Reinsurance Constrained Multi-Agent Simulation Process (R-CMASP)LLM agents and tool use → Financial trading and forecasting agents
PrefGen: Multimodal Preference Learning for Preference-Conditioned Image GenerationLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Enhancing Deep Deterministic Policy Gradients on Continuous Control Tasks with Decoupled Prioritized Experience ReplayDeep reinforcement learning methods → Deep RL algorithms and representations
To Think or Not to Think: The Hidden Cost of Meta-Training with Excessive CoT ExamplesLLM reasoning → Prompting and chain-of-thought reasoning
The Erosion of LLM Signatures: Can We Still Distinguish Human and LLM-Generated Scientific Ideas After Iterative Paraphrasing?LLM agents and tool use → Web, search, and research agents
CFO: Learning Continuous-Time PDE Dynamics via Flow-Matched Neural OperatorsDiffusion models → Flow matching, time series and science
SHAP-Guided Kernel Actor-Critic for Explainable Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
From Segments to Scenes: Temporal Understanding for Agentic Autonomous Driving via Vision-Language ModelsMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection +2
CARD: Correlation Aware Restoration with DiffusionDiffusion models → Inverse problems and image restoration
When unlearning is free: leveraging low influence points to reduce computational costsLLM safety and security → Machine unlearning
One-Step Diffusion Samplers via Self-Distillation and Deterministic FlowDiffusion models → Diffusion foundations and sampling
IE2Video: Adapting Pretrained Diffusion Models for Event-Based Video ReconstructionDiffusion models → Video generation
Invariance Co-training for Robot Visual GeneralizationVision-language-action models → World model-based VLA
Rethinking Tokenization for Clinical Time Series: When Less is MoreLLM reasoning → Clinical and biomedical applications
Your Latent Mask is Wrong: Pixel-Equivalent Latent Compositing for Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Semore: VLM-guided Enhanced Semantic Motion Representations for Visual Reinforcement LearningMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Value Gradient Guidance for Flow Matching AlignmentDiffusion models → Flow matching, time series and science +1
DraCo: Draft as CoT for Text-to-Image Preview and Rare Concept GenerationLLM reasoning → Multimodal reasoning +1
ARM-Thinker: Reinforcing Multimodal Generative Reward Models with Agentic Tool Use and Visual ReasoningLLM reasoning → Multimodal reasoning +2
STARE-VLA: Progressive Stage-Aware Reinforcement for Fine-Tuning Vision-Language-Action ModelsVision-language-action models → Reinforcement learning and post-training
Semantic Soft Bootstrapping: Long Context Reasoning in LLMs without Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning +2
Structured Document Translation via Format Reinforcement LearningLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Foundations of Diffusion Models in General State Spaces: A Self-Contained IntroductionDiffusion models → Diffusion foundations and sampling
Visual Reasoning Tracer: Object-Level Grounded Reasoning BenchmarkLLM reasoning → Multimodal reasoning +1
Deep Forcing: Training-Free Long Video Generation with Deep Sink and Participative CompressionDiffusion models → Video generation +1
BulletTime: Decoupled Control of Time and Camera Pose for Video GenerationDiffusion models → Video generation
David vs. Goliath: Can Small Models Win Big with Agentic AI in Hardware Design?LLM agents and tool use → Software engineering coding agents
Personalizing Agent Privacy Decisions via Logical EntailmentLLM agents and tool use → Agent threats, guardrails and governance
Arbitrage: Efficient Reasoning via Advantage-Aware SpeculationLLM reasoning → Test-time compute and efficient thinking +1
Generative Neural Video Compression via Video Diffusion PriorDiffusion models → Diffusion for compression and wireless
Reflection Removal through Efficient Adaptation of Diffusion TransformersDiffusion models → Inverse problems and image restoration
When AI Agents Compete for Jobs: Strategic Capabilities and Economic Dynamics of AI Labour MarketsLLM agents and tool use → Social simulation and multi-agent debate
Nex-N1: Agentic Models Trained via a Unified Ecosystem for Large-Scale Environment ConstructionLLM agents and tool use → General LLM agent frameworks and benchmarks
Towards a unified framework for guided diffusion modelsDiffusion models → Diffusion foundations and sampling
Environment-Aware Channel Inference via Cross-Modal Flow: From Multimodal Sensing to Wireless ChannelsDiffusion models → Diffusion for compression and wireless
Hybrid Imitation Learning: Teleoperation Augmentation Primitives that Policies Learn to TriggerOffline RL and imitation learning from data → Robot manipulation imitation learning
Realizable Abstractions: Near-Optimal Hierarchical Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
FASTer: Toward Efficient Autoregressive Vision Language Action Modeling via Neural Action TokenizationVision-language-action models → Efficient VLA inference and adaptation
CARL: Criticality-Aware Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents
LiteVGGT: Boosting Vanilla VGGT via Geometry-aware Cached Token MergingEfficient LLM inference and serving → Token pruning and merging for ViTs
Semantics Lead the Way: Harmonizing Semantic and Texture Modeling with Asynchronous Latent DiffusionDiffusion models → Image generation, text-to-image and editing
Algorithmic Thinking TheoryLLM reasoning → Reasoning benchmarks and evaluation
ReflexFlow: Rethinking Learning Objective for Exposure Bias Alleviation in Flow MatchingDiffusion models → Flow matching, time series and science
Chameleon: Adaptive Adversarial Agents for Scaling-Based Visual Prompt Injection in Multimodal AI SystemsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +2
STELLA: Guiding Large Language Models for Time Series Forecasting with Semantic AbstractionsLLM reasoning → Time series reasoning
SEAL: Self-Evolving Agentic Learning for Conversational Question Answering over Knowledge GraphsLLM reasoning → Knowledge- and tool-augmented reasoning +2
Are Your Agents Upward Deceivers?LLM agents and tool use → Social simulation and multi-agent debate
Autoregressive Image Generation Needs Only a Few Lines of Cached TokensEfficient LLM inference and serving → Fast autoregressive visual generation
Safe model-based Reinforcement Learning via Model Predictive Control and Control Barrier FunctionsDeep reinforcement learning methods → Safe and constrained RL
SoK: a Comprehensive Causality Analysis Framework for Large Language Model SecurityLLM safety and security → Jailbreaking
FreeGen: Feed-Forward Reconstruction-Generation Co-Training for Free-Viewpoint Driving Scene SynthesisEnd-to-end autonomous driving → Generative simulation for driving
Enabling Ethical AI: A case study in using Ontological Context for Justified Agentic AI DecisionsLLM agents and tool use → General LLM agent frameworks and benchmarks
LatentFM: A Latent Flow Matching Approach for Generative Medical Image SegmentationDiffusion models → Medical imaging and segmentation
MOVE: A Simple Motion-Based Data Collection Paradigm for Spatial Generalization in Robotic ManipulationOffline RL and imitation learning from data → Robot manipulation imitation learning
SIMA 2: A Generalist Embodied Agent for Virtual WorldsLLM agents and tool use → Reinforcement learning for LLM agents
Spatially-Enhanced Retrieval-Augmented Generation for Walkability and Urban DiscoveryRetrieval-augmented generation → RAG for recommendation
ASTRIDE: A Security Threat Modeling Platform for Agentic-AI ApplicationsLLM agents and tool use → Agent threats, guardrails and governance +1
Complementary Characterization of Agent-Based Models via Computational Mechanics and Diffusion ModelsDiffusion models → Diffusion foundations and sampling
RLHFSpec: Breaking the Efficiency Bottleneck in RLHF Training via Adaptive DraftingLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
UnwrapDiff: A Conditional Diffusion Model for InSAR Phase UnwrappingDiffusion models → Inverse problems and image restoration
Model Whisper: Steering Vectors Unlock Large Language Models' Potential in Test-timeLLM reasoning → Test-time compute and efficient thinking +1
E3AD: An Emotion-Aware Vision-Language-Action Model for Human-Centric End-to-End Autonomous DrivingEnd-to-end autonomous driving → Reasoning VLA and VLM driving policies
Measuring the Unspoken: A Disentanglement Model and Benchmark for Psychological Analysis in the WildMultimodal LLMs / vision-language models → Multimodal emotion understanding
M3-TTS: Multi-modal DiT Alignment & Mel-latent for Zero-shot High-fidelity Speech SynthesisDiffusion models → Audio, speech and music
POLARIS: Is Multi-Agentic Reasoning the Next Wave in Engineering Self-Adaptive Systems?LLM agents and tool use → General LLM agent frameworks and benchmarks
OmniScaleSR: Unleashing Scale-Controlled Diffusion Prior for Faithful and Realistic Arbitrary-Scale Image Super-ResolutionDiffusion models → Inverse problems and image restoration
Continuous-time reinforcement learning for optimal switching over multiple regimesDeep reinforcement learning methods → RL theory and sample complexity
Towards Ethical Multi-Agent Systems of Large Language Models: A Mechanistic Interpretability PerspectiveLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Towards Cross-View Point Correspondence in Vision-Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite LengthDiffusion models → Audio, speech and music
Topology Matters: Measuring Memory Leakage in Multi-Agent LLMsLLM agents and tool use → Agent threats, guardrails and governance
Rethinking Decoupled Knowledge Distillation: A Predictive Distribution PerspectiveKnowledge distillation → General knowledge distillation methods
Denoise to Track: Harnessing Video Diffusion Priors for Robust CorrespondenceDiffusion models → Video generation
PBFuzz: Agentic Directed Fuzzing for PoV GenerationLLM agents and tool use → Agent threats, guardrails and governance
Natural Language Actor-Critic: Scalable Off-Policy Learning in Language SpaceLLM reasoning → Reasoning for agents, planning and code
QoSDiff: An Implicit Topological Embedding Learning Framework Leveraging Denoising Diffusion and Adversarial Attention for Robust QoS PredictionDiffusion models → Flow matching, time series and science
COOPER: A Unified Model for Cooperative Perception and Reasoning in Spatial IntelligenceLLM reasoning → Spatial and embodied reasoning +1
Omniscient Attacker in Stochastic Security Games with Interdependent NodesDeep reinforcement learning methods → Adversarial robustness and RL in finance
Diffusion Fine-Tuning via Reparameterized Policy Gradient of the Soft Q-FunctionDiffusion models → Diffusion and flow policies for robotics +1
On the Limits of Test-Time Compute: Sequential Reward Filtering for Better InferenceLLM reasoning → Test-time compute and efficient thinking
Efficient Safety Verification of Autonomous Vehicles with Neural Network OperatorEnd-to-end autonomous driving → End-to-end system architecture
RRPO: Robust Reward Policy Optimization for LLM-based Emotional TTSLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
AdmTree: Compressing Lengthy Context with Adaptive Semantic TreesEfficient LLM inference and serving → Prompt and context compression
VideoMem: Enhancing Ultra-Long Video Understanding via Adaptive Memory ManagementMultimodal LLMs / vision-language models → Video understanding with MLLMs
GTM: Simulating the World of Tools for AI AgentsLLM agents and tool use → Reinforcement learning for LLM agents
Refaccade: Editing Object with Given Reference TextureDiffusion models → 3D, point cloud and scene generation
BiTAgent: A Task-Aware Modular Framework for Bidirectional Coupling between Multimodal Large Language Models and World ModelsWorld models → World models for language agents +1
UltraImage: Rethinking Resolution Extrapolation in Image Diffusion TransformersDiffusion models → Diffusion foundations and sampling
Back to Basics: Motion Representation Matters for Human Motion Generation Using Diffusion ModelDiffusion models → Video generation
Context-Aware Mixture-of-Experts Inference on CXL-Enabled GPU-NDP SystemsMixture-of-experts → Efficient MoE training and inference systems
UniTS: Unified Spatio-Temporal Generative Model for Remote SensingDiffusion models → Inverse problems and image restoration
dVLM-AD: Enhance Diffusion Vision-Language-Model for Driving via Controllable ReasoningLLM reasoning → Chain-of-thought reasoning for driving +2
RapidUn: Influence-Driven Parameter Reweighting for Efficient Large Language Model UnlearningLLM safety and security → Machine unlearning
GuidNoise: Single-Pair Guided Diffusion for Generalized Noise SynthesisDiffusion models → Diffusion foundations and sampling
Vision-Language-Action Models for Selective Robotic Disassembly: A Case Study on Critical Component Extraction from DesktopsVision-language-action models → Generalist VLA models and benchmarks
MindDrive: An All-in-One Framework Bridging World Models and Vision-Language Model for End-to-End Autonomous DrivingWorld models → World models for autonomous driving +1
Quantum-Accelerated Deep Reinforcement Learning for Frequency Regulation EnhancementDeep reinforcement learning methods → Quantum reinforcement learning
DataGovBench: Benchmarking LLM Agents for Real-World Data Governance WorkflowsLLM agents and tool use → Web, search, and research agents

All 351