Papers

New on arXiv

Sun 5 Oct 2025

235 papers

Systematic Diagnosis of Brittle Reasoning in Large Language ModelsLLM reasoning → Mathematical reasoning
AlphaApollo: A System for Deep Agentic ReasoningLLM agents and tool use → General LLM agent frameworks and benchmarks
Agentic Misalignment: How LLMs Could Be Insider ThreatsLLM agents and tool use → Agent threats, guardrails and governance
PatternKV: Flattening KV Representation Expands Quantization HeadroomLLM reasoning → Test-time compute and efficient thinking +1
Emergent Coordination in Multi-Agent Language ModelsLLM agents and tool use → Social simulation and multi-agent debate
SafeGuider: Robust and Practical Content Safety Control for Text-to-Image ModelsLLM safety and security → Vision-language and text-to-image safety
From Poisoned to Aware: Fostering Backdoor Self-Awareness in LLMsLLM safety and security → Backdoor and data poisoning attacks
On The Statistical Limits of Self-Improving AgentsSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
Improving Consistency in Retrieval-Augmented Systems with Group Similarity RewardsRetrieval-augmented generation → RAG robustness and adaptive retrieval
MorphoSim: An Interactive, Controllable, and Editable Language-guided 4D World SimulatorWorld models → Embodied and robot world models
JEF-Hinter: Leveraging Offline Knowledge for Improving Web Agents AdaptationLLM agents and tool use → Web, search, and research agents
Speculative Actions: A Lossless Framework for Faster Agentic SystemsLLM agents and tool use → Software engineering coding agents
Diffusion^2: Dual Diffusion Model with Uncertainty-Aware Adaptive Noise for Momentary Trajectory PredictionDiffusion models → Diffusion and flow policies for robotics
Reflection Before Action: Designing a Framework for Quantifying Thought Patterns for Increased Self-awareness in Personal Decision MakingLLM reasoning → Reasoning benchmarks and evaluation
Quantizer Design for Finite Model Approximations, Model Learning, and Quantized Q-Learning for MDPs with Unbounded SpacesDeep reinforcement learning methods → RL theory and sample complexity
Reliable and Scalable Robot Policy Evaluation with Imperfect SimulatorsOffline RL and imitation learning from data → Robot manipulation imitation learning
Unmasking Backdoors: An Explainable Defense via Gradient-Attention Anomaly Scoring for Pre-trained Language ModelsLLM safety and security → Backdoor and data poisoning attacks
RAP: 3D Rasterization Augmented End-to-End PlanningEnd-to-end autonomous driving → Imitation-learning end-to-end planners +1
FoilDiff: A Hybrid Transformer Backbone for Diffusion-based Modelling of 2D Airfoil Flow FieldsDiffusion models → Flow matching, time series and science
Read the Scene, Not the Script: Outcome-Aware Safety for LLMsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
FairAgent: Democratizing Fairness-Aware Machine Learning with LLM-Powered AgentsLLM agents and tool use → Social simulation and multi-agent debate
On the Importance of Task Complexity in Evaluating LLM-Based Multi-Agent SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
Activation Steering with a Feedback ControllerLLM alignment / RLHF / preference optimisation → LLM preference optimization methods +2
Audit the Whisper: Detecting Steganographic Collusion in Multi-Agent LLMsLLM agents and tool use → Agent threats, guardrails and governance
Equipping Retrieval-Augmented Large Language Models with Document Structure AwarenessRetrieval-augmented generation → Multimodal RAG
ChronoEdit: Towards Temporal Reasoning for Image Editing and World SimulationLLM reasoning → Multimodal reasoning
SliceMoE: Routing Embedding Slices Instead of Tokens for Fine-Grained and Balanced Transformer ScalingMixture-of-experts → Expert routing strategies
Doctor-R1: Mastering Clinical Inquiry with Experiential Agentic Reinforcement LearningLLM agents and tool use → Medical and scientific domain agents
RetiBridge: Bridging Quantitative Retinal Biomarkers and Qualitative Diagnosis with a Knowledge-Guided Multimodal Large Language ModelMultimodal LLMs / vision-language models → Medical vision-language models
A KL-regularization Framework for Learning to Plan with Adaptive PriorsDeep reinforcement learning methods → RL theory and sample complexity
Closing the Loop: Coordinating Inventory and Recommendation via Deep Reinforcement Learning on Multiple TimescalesDeep reinforcement learning methods → RL for energy, networks and operations
Small Fleet, Big Impact: Enhancing Shared Micromobility Efficiency through Minimal Autonomous Vehicle DeploymentEnd-to-end autonomous driving → Cooperative and V2X driving
Don't Pass@k: A Bayesian Framework for Large Language Model EvaluationLLM reasoning → Test-time compute and efficient thinking
VortexPIA: Indirect Prompt Injection Attack against LLMs for Efficient Extraction of User PrivacyLLM safety and security → Prompt injection and agent security
AgentTypo: Adaptive Typographic Prompt Injection Attacks against Black-box Multimodal AgentsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +2
ContextVLA: Vision-Language-Action Model with Amortized Multi-Frame ContextVision-language-action models → Generalist VLA models and benchmarks
Flexible Locomotion Learning with Diffusion Model Predictive ControlDiffusion models → Diffusion and flow policies for robotics
Pushing on Multilingual Reasoning Models with Language-Mixed Chain-of-ThoughtLLM reasoning → Reasoning benchmarks and evaluation
Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated ImagesLLM reasoning → Multimodal reasoning
MLLMEraser: Achieving Test-Time Unlearning in Multimodal Large Language Models through Activation SteeringMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +2
Teaching LLM to be Persuasive: Reward-Enhanced Policy Optimization for Alignment from Heterogeneous RewardsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Enhancing Speaker Verification with w2v-BERT 2.0 and Knowledge Distillation guided Structured PruningKnowledge distillation → Speech and audio distillation
AgentRL: Scaling Agentic Reinforcement Learning with a Multi-Turn, Multi-Task FrameworkLLM agents and tool use → Reinforcement learning for LLM agents
CALM Before the STORM: Unlocking Native Reasoning for Optimization ModelingLLM reasoning → Test-time compute and efficient thinking
COSMO-RL: Towards Trustworthy LMRMs via Joint Safety and StabilityLLM reasoning → Reasoning for agents, planning and code +1
Constructing coherent spatial memory in LLM agents through graph rectificationLLM agents and tool use → Agent memory systems
Finite Time Analysis of Constrained Natural Critic-Actor Algorithm with Improved Sample ComplexityDeep reinforcement learning methods → RL theory and sample complexity
Let Features Decide Their Own Solvers: Hybrid Feature Caching for Diffusion TransformersDiffusion models → Diffusion foundations and sampling
Thinking on the Fly: Test-Time Reasoning Enhancement via Latent Thought Policy OptimizationLLM reasoning → Test-time compute and efficient thinking
Open Agent Specification (Agent Spec): A Unified Representation for AI AgentsLLM agents and tool use → Software engineering coding agents
Drax: Speech Recognition with Discrete Flow MatchingDiffusion models → Audio, speech and music
GDiffuSE: Diffusion-based speech enhancement with noise model guidanceDiffusion models → Audio, speech and music
ObCLIP: Oblivious CLoud-Device Hybrid Image Generation with Privacy PreservationDiffusion models → Robustness, privacy and adversarial aspects
Self Speculative Decoding for Diffusion Large Language ModelsEfficient LLM inference and serving → Speculative decoding
Automating construction safety inspections using a multi-modal vision-language RAG frameworkRetrieval-augmented generation → Agentic RAG, agents and memory
Turning Drift into Constraint: Robust Reasoning Alignment in Non-Stationary Multi-Stream EnvironmentsLLM reasoning → Test-time compute and efficient thinking +2
The Artificial Intelligence Cognitive Examination: A Survey on the Evolution of Multimodal Evaluation from Recognition to ReasoningLLM reasoning → Multimodal reasoning
Selective Expert Guidance for Effective and Diverse Exploration in Reinforcement Learning of LLMsLLM reasoning → Reinforcement learning for reasoning
MoME: Mixture of Matryoshka Experts for Audio-Visual Speech RecognitionMixture-of-experts → Multimodal MoE models
GA4GC: Greener Agent for Greener Code via Multi-Objective Configuration OptimizationLLM agents and tool use → Software engineering coding agents
Internal states before wait modulate reasoning patternsLLM reasoning → Test-time compute and efficient thinking
Projection and Quantisation: A Unifying View of Learning to Hash, from Random Projections to the RAG EraRetrieval-augmented generation → Generation and context integration
Joint Learning of Pose Regression and Denoising Diffusion with Score Scaling Sampling for Category-level 6D Pose EstimationDiffusion models → 3D, point cloud and scene generation
Searching Meta Reasoning Skeleton to Guide LLM ReasoningLLM reasoning → Test-time compute and efficient thinking
Can Linear Probes Measure LLM Uncertainty?Mechanistic interpretability → SAE features and their applications
RLRF: Competitive Search Agent Design via Reinforcement Learning from Ranker FeedbackLLM agents and tool use → Web, search, and research agents +2
Harnessing LLM for Noise-Robust Cognitive Diagnosis in Web-Based Intelligent Education SystemsDiffusion models → Discrete and text diffusion
SPOGW: a Score-based Preference Optimization method via Group-Wise comparison for workflowsLLM agents and tool use → Reinforcement learning for LLM agents +1
Offline Reinforcement Learning in Large State Spaces: Algorithms and GuaranteesOffline RL and imitation learning from data → Offline RL algorithms
Best of mini-N in-loop Sampling: A Contextual Quality Reward Model for Reliable and Efficient Best-of-N SamplingLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Scaling Code-Assisted Chain-of-Thoughts and Instructions for Model ReasoningLLM reasoning → Prompting and chain-of-thought reasoning
PoLi-RL: A Point-to-List Reinforcement Learning Framework for Conditional Semantic Textual SimilarityDeep reinforcement learning methods → RL training of language and agent models
From Shadow to Light: Toward Safe and Efficient Policy Learning Across MPC, DeePC, RL, and LLM AgentsLLM agents and tool use → Reinforcement learning for LLM agents
Moral Anchor System: A Predictive Framework for AI Value Alignment and Drift PreventionLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Slow-Fast Policy Optimization: Reposition-Before-Update for LLM ReasoningLLM reasoning → Reinforcement learning for reasoning
What Makes Diffusion Language Models Super Data Learners?Diffusion models → Discrete and text diffusion
Unlearning in Diffusion models under Data Constraints: A Variational Inference ApproachDiffusion models → Robustness, privacy and adversarial aspects +1
Exploring Chain-of-Thought Reasoning for Steerable Pluralistic AlignmentLLM reasoning → Prompting and chain-of-thought reasoning
SITCOM: Scaling Inference-Time COMpute for VLAsLLM reasoning → Spatial and embodied reasoning
FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought ReasoningLLM reasoning → Prompting and chain-of-thought reasoning
textscGUI-Spotlight: Adaptive Iterative Focus Refinement for Enhanced GUI Visual GroundingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Prompt-to-Prompt: Text-Based Image Editing Via Cross-Attention Mechanisms -- The Research of Hyperparameters and Novel Mechanisms to Enhance Existing FrameworksDiffusion models → Image generation, text-to-image and editing
The Debate on RLVR Reasoning Capability Boundary: Shrinkage, Expansion, or Both? A Two-Stage Dynamic ViewLLM reasoning → Reinforcement learning for reasoning
LLM-Based Data Science Agents: A Survey of Capabilities, Challenges, and Future DirectionsLLM reasoning → Reasoning for agents, planning and code +1
Spatiotemporal Forecasting as Planning: A Model-Based Reinforcement Learning Approach with Generative World ModelsWorld models → Latent predictive world models +1
Simple Policy Gradients for Reasoning with Diffusion Language ModelsLLM reasoning → Diffusion reasoning and test-time scaling +1
Zephyrus: An Agentic Framework for Weather ScienceLLM agents and tool use → Financial trading and forecasting agents
Thai Semantic End-of-Turn Detection for Real-Time Voice AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
AgriGPT-VL: Agricultural Vision-Language Understanding SuiteLLM reasoning → Clinical and biomedical applications +1
Quantitative Certification of Agentic Tool SelectionLLM agents and tool use → Tool learning and function calling
The Signal is in the Steps: Local Scoring for Reasoning Data SelectionKnowledge distillation → Knowledge distillation for LLMs
Beyond Static Evaluation: Rethinking the Assessment of Personalized Agent Adaptability in Information RetrievalLLM reasoning → Knowledge- and tool-augmented reasoning +1
Enhancing Talent Search Ranking with Role-Aware Expert Mixtures and LLM-based Fine-Grained Job Descriptions
Benchmarking Open-Source Large Language Models for Persian in Zero-Shot and Few-Shot Learning
Rounding-Guided Backdoor Injection in Deep Learning Model Quantization
Real-Time Health Analytics Using Ontology-Driven Complex Event Processing and LLM Reasoning: A Tuberculosis Case Study
Confidence, Not Perplexity: A Better Metric for the Creative Era of LLMs
Hierarchical Self-Supervised Representation Learning for Depression Detection from Speech
Large Language Models Hallucination: A Comprehensive Survey
A Mixed-Methods Analysis of Repression and Mobilization in Bangladesh's July Revolution Using Machine Learning and Statistical Modeling

All 235