Papers

New on arXiv

Mon 1 Dec 2025

413 papers

Dynamic System Instructions and Tool Exposure for Efficient Agentic LLMsLLM agents and tool use → Tool learning and function calling
A Comparative Study of Student Perspectives on Technical Writing Feedback Quality: Evaluating LLMs, SLMs, and Humans in Computer Science TopicsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
KVReviver: Reversible KV Cache Compression with Sketch-Based Token ReconstructionEfficient LLM inference and serving → KV cache optimization
StructuredDNA: A Bio-Physical Framework for Energy-Aware Transformer RoutingMixture-of-experts → Expert routing strategies
CluCERT: Certifying LLM Robustness via Clustering-Guided Denoising SmoothingLLM safety and security → Jailbreaking
FlowEO: Generative Unsupervised Domain Adaptation for Earth ObservationDiffusion models → Inverse problems and image restoration
Spatiotemporal Satellite Image Downscaling with Transfer Encoders and Autoregressive Generative ModelsDiffusion models → Inverse problems and image restoration
ALARM: Automated MLLM-Based Anomaly Detection in Complex-EnviRonment Monitoring with Uncertainty QuantificationMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Ensemble Privacy Defense for Knowledge-Intensive LLMs against Membership Inference AttacksRetrieval-augmented generation → Privacy and security of RAG +1
Many-to-One Adversarial Consensus: Exposing Multi-Agent Collusion Risks in AI-Based HealthcareLLM agents and tool use → Medical and scientific domain agents
PhyCustom: Towards Realistic Physical Customization in Text-to-Image GenerationDiffusion models → Image generation, text-to-image and editing
IC-World: In-Context Generation for Shared World ModelingWorld models → Video-generation world models and simulators
DialogGuard: Multi-Agent Psychosocial Safety Evaluation of Sensitive LLM ResponsesLLM agents and tool use → Medical and scientific domain agents
Spatiotemporal Pyramid Flow Matching for Climate EmulationDiffusion models → Flow matching, time series and science
TradeTrap: Are LLM-based Trading Agents Truly Reliable and Faithful?LLM agents and tool use → Financial trading and forecasting agents
DETAIL Matters: Measuring the Impact of Prompt Specificity on Reasoning in Large Language ModelsLLM reasoning → Prompting and chain-of-thought reasoning
Lightweight Latent Reasoning for Narrative TasksLLM reasoning → Test-time compute and efficient thinking
See, Hear, and Understand: Benchmarking Audiovisual Human Speech Understanding in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Video understanding with MLLMs
Benchmarking LLM Agents for Wealth-Management WorkflowsLLM agents and tool use → Financial trading and forecasting agents
STRIDE: A Systematic Framework for Selecting AI Modalities -- Agentic AI, AI Assistants, or LLM CallsLLM agents and tool use → General LLM agent frameworks and benchmarks
Orchestration Framework for Financial Agents: From Algorithmic Trading to Agentic TradingLLM agents and tool use → Financial trading and forecasting agents
Towards Unified Video Quality AssessmentMixture-of-experts → Vision mixture-of-experts
Improved Training Mechanism for Reinforcement Learning via Online Model SelectionDeep reinforcement learning methods → Deep RL algorithms and representations
A Knowledge-Based Language Model: Deducing Grammatical Knowledge in a Multi-Agent Language Acquisition SimulationLLM agents and tool use → General LLM agent frameworks and benchmarks
Enforcing Orderedness to Improve Feature ConsistencyMechanistic interpretability → Sparse autoencoder training and evaluation
From monoliths to modules: Decomposing transducers for efficient world modellingWorld models → World models for language agents
Think Before You Prune: Self-Reflective Structured Pruning for Reasoning Language ModelsLLM reasoning → Test-time compute and efficient thinking
FineGRAIN: Evaluating Failure Modes of Text-to-Image Models with Vision Language Model JudgesMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Factor(T,U): Factored Cognition Strengthens Monitoring of Untrusted AILLM safety and security → Prompt injection and agent security
Diffusion-Augmented Markov Decision Processes for Maximum Entropy Reinforcement LearningDiffusion models → Diffusion and flow policies for robotics +1
Objects in Generated Videos Are Slower Than They Appear: Models Suffer Sub-Earth Gravity and Don't Know Galileo's Principle...for nowWorld models → Latent predictive world models
ManualVLA: A Unified VLA Model for Chain-of-Thought Manual Generation and Robotic ManipulationLLM reasoning → Spatial and embodied reasoning +1
Learning Dexterous Manipulation Skills from Imperfect SimulationsDeep reinforcement learning methods → RL for robot locomotion and manipulation
The Art of Scaling Test-Time Compute for Large Language ModelsLLM reasoning → Test-time compute and efficient thinking
AlignSAE: Concept-Aligned Sparse AutoencodersMechanistic interpretability → SAE features and their applications
Learning Sim-to-Real Humanoid Locomotion in 15 MinutesDeep reinforcement learning methods → RL for robot locomotion and manipulation
RoaD: Rollouts as Demonstrations for Closed-Loop Supervised Fine-Tuning of Autonomous Driving PoliciesEnd-to-end autonomous driving → Imitation-learning end-to-end planners
LLM CHESS: Benchmarking Reasoning and Instruction-Following in LLMs through ChessLLM reasoning → Reasoning for agents, planning and code +1
PAI-Bench: A Comprehensive Benchmark For Physical AIMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Artemis: Structured Visual Reasoning for Perception Policy LearningLLM reasoning → Multimodal reasoning
Forecasting in Offline Reinforcement Learning for Non-stationary EnvironmentsOffline RL and imitation learning from data → Offline RL algorithms
Chain-of-Ground: Improving GUI Grounding via Iterative Reasoning and Reference FeedbackLLM reasoning → Multimodal reasoning +1
SGDiff: Scene Graph Guided Diffusion Model for Image Collaborative SegCaptioningDiffusion models → Image generation, text-to-image and editing
KV Pareto: Systems-Level Optimization of KV Cache and Model Compression for Long Context InferenceEfficient LLM inference and serving → KV cache optimization
GrndCtrl: Grounding World Models via Self-Supervised Reward AlignmentWorld models → Latent dynamics models and model-based RL
Script: Graph-Structured and Query-Conditioned Semantic Token Pruning for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
How Far Are We from Genuinely Useful Deep Research Agents?LLM agents and tool use → Web, search, and research agents
Guardian: Detecting Robotic Planning and Execution Errors with Vision-Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Agentic Policy Optimization via Instruction-Policy Co-EvolutionLLM reasoning → Reinforcement learning for reasoning +1
An Empirical Study of Agent Developer Practices in AI Agent FrameworksLLM agents and tool use → Software engineering coding agents
Rectifying LLM Thought from Lens of OptimizationLLM reasoning → Test-time compute and efficient thinking
Real-World Robot Control by Deep Active Inference With a Temporally Hierarchical World ModelWorld models → Embodied and robot world models
Med-VCD: Mitigating Hallucination for Medical Large Vision Language Models through Visual Contrastive DecodingMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Latent Debate: A Surrogate Framework for Interpreting LLM ThinkingLLM reasoning → Hallucination detection and mitigation
Exploring Human Perceptions of AI Responses: Insights from a Mixed-Methods Study on Risk Mitigation in Generative ModelsLLM safety and security → Safety alignment, refusal and guardrails
New Spiking Architecture for Multi-Modal Decision-Making in Autonomous VehiclesEnd-to-end autonomous driving → End-to-end system architecture +1
Register Any Point: Scaling 3D Point Cloud Registration by Flow MatchingDiffusion models → 3D, point cloud and scene generation
Beyond SFT: Reinforcement Learning for Safer Large Reasoning Models with Better Reasoning AbilityLLM reasoning → Reasoning for agents, planning and code +1
OpenREAD: Reinforced Open-Ended Reasoning for End-to-End Autonomous Driving with LLM-as-CriticLLM reasoning → Chain-of-thought reasoning for driving +1
InnoGym: Benchmarking the Innovation Potential of AI AgentsLLM reasoning → Reasoning for agents, planning and code +1
Seeing through Imagination: Learning Scene Geometry via Implicit Spatial World ModelingWorld models → World models for language agents +2
Dimension-free error estimate for diffusion model and optimal schedulingDiffusion models → Diffusion foundations and sampling
GR-RL: Going Dexterous and Precise for Long-Horizon Robotic ManipulationVision-language-action models → Reinforcement learning and post-training
How Does RL Post-training Induce Skill Composition? A Case Study on CountdownLLM reasoning → Reinforcement learning for reasoning
IGen: Scalable Data Generation for Robot Learning from Open-World ImagesVision-language-action models → World model-based VLA
HiconAgent: History Context-aware Policy Optimization for GUI AgentsLLM agents and tool use → GUI and mobile agents
Multimodal Mixture-of-Experts for ISAC in Low-Altitude Wireless NetworksMixture-of-experts → Expert routing strategies
Automating modeling in mechanics: LLMs as designers of physics-constrained neural networks for constitutive modeling of materialsLLM agents and tool use → Multimodal and embodied agents
Beware of Reasoning Overconfidence: Pitfalls in the Reasoning Process for Multi-solution TasksLLM reasoning → Test-time compute and efficient thinking
Probabilistic Neuro-Symbolic Reasoning for Sparse Historical Data: A Framework Integrating Bayesian Inference, Causal Models, and Game-Theoretic AllocationLLM reasoning → Causal reasoning in LLMs
Transport Discrepancy as a Reliability Signal for Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation +1
StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming VideosLLM reasoning → Multimodal reasoning
GRASP: Guided Residual Adapters with Sample-wise PartitioningDiffusion models → Flow matching, time series and science
Bridging the Scale Gap: Balanced Tiny and General Object Detection in Remote Sensing ImageryMixture-of-experts → Vision mixture-of-experts
HalluGraph: Auditable Hallucination Detection for Legal RAG Systems via Knowledge Graph AlignmentRetrieval-augmented generation → Hallucination detection in RAG
Agent-Kernel: A MicroKernel Multi-Agent System Framework for Adaptive Social Simulation Powered by LLMsLLM agents and tool use → Social simulation and multi-agent debate
Reconstructing Multi-Scale Physical Fields from Extremely Sparse Measurements with an Autoencoder-Diffusion CascadeDiffusion models → Inverse problems and image restoration
MasHeNe: A Benchmark for Head and Neck CT Mass Segmentation using Window-Enhanced Mamba with Frequency-Domain IntegrationState-space sequence models (S4, Mamba) → Mamba for medical image analysis
NavForesee: A Unified Vision-Language World Model for Hierarchical Planning and Dual-Horizon Navigation PredictionWorld models → World models for navigation
Diffusion Fuzzy System: Fuzzy Rule Guided Latent Multi-Path Diffusion ModelingDiffusion models → Diffusion foundations and sampling
End-to-end Deep Reinforcement Learning for Stochastic Multi-objective Optimization in C-VRPTWDeep reinforcement learning methods → RL for games and combinatorial optimization
MCAT: Scaling Many-to-Many Speech-to-Text Translation with MLLMs to 70 LanguagesMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Formal Verification of Noisy Quantum Reinforcement Learning PoliciesDeep reinforcement learning methods → Quantum reinforcement learning
No Trust Issues Here: A Technical Report on the Winning Solutions for the Rayan AI ContestLLM safety and security → Backdoor and data poisoning attacks
Multi-Path Collaborative Reasoning via Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning
ChronosObserver: Taming 4D World with Hyperspace Diffusion SamplingDiffusion models → 3D, point cloud and scene generation
Zero-Overhead Introspection for Adaptive Test-Time ComputeLLM reasoning → Test-time compute and efficient thinking
A Flexible Multi-Agent LLM-Human Framework for Fast Human Validated Tool BuildingLLM agents and tool use → Tool learning and function calling
ResDiT: Evoking the Intrinsic Resolution Scalability in Diffusion TransformersDiffusion models → Diffusion foundations and sampling
ViRectify: A Challenging Benchmark for Video Reasoning Correction with Multimodal Large Language ModelsLLM reasoning → Multimodal reasoning +1
MDiff4STR: Mask Diffusion Model for Scene Text RecognitionDiffusion models → Discrete and text diffusion
PromptBridge: Cross-Model Prompt Transfer for Large Language ModelsLLM reasoning → Prompting and chain-of-thought reasoning
Rice-VL: Evaluating Vision-Language Models for Cultural Understanding Across ASEAN CountriesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
FRAMER: Frequency-Aligned Self-Distillation with Adaptive Modulation Leveraging Diffusion Priors for Real-World Image Super-ResolutionDiffusion models → Inverse problems and image restoration +1
Syndrome-Flow Consistency Model Achieves One-step Denoising Error Correction CodesDiffusion models → Discrete and text diffusion
PointNet4D: A Lightweight 4D Point Cloud Video Backbone for Online and Offline Perception in Robotic ApplicationsState-space sequence models (S4, Mamba) → Point clouds, detection and remote sensing
Stabilizing Reinforcement Learning with LLMs: Formulation and PracticesMixture-of-experts → MoE in reinforcement learning and robotics
LAURA: Enhancing Code Review Generation with Context-Enriched Retrieval-Augmented LLMRetrieval-augmented generation → RAG for code
The Necessity of Imperfection:Reversing Model Collapse via Simulating Cognitive BoundednessSynthetic data and self-improvement for LLMs → Model collapse from synthetic data
The Trojan Knowledge: Bypassing Commercial LLM Guardrails via Harmless Prompt Weaving and Adaptive Tree SearchLLM safety and security → Jailbreaking

All 413