Papers

New on arXiv

Thu 18 Sep 2025

404 papers

Semantic Representation Attack against Aligned Large Language ModelsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
VLA-LPAF: Lightweight Perspective-Adaptive Fusion for Vision-Language-Action to Enable More Unconstrained Robotic ManipulationVision-language-action models → Generalist VLA models and benchmarks
SecureFixAgent: A Hybrid LLM Agent for Automated Python Static Vulnerability RepairLLM agents and tool use → Agent threats, guardrails and governance
Digging Into the Internal: Causality-Based Analysis of LLM Function CallingLLM agents and tool use → Tool learning and function calling
SmolRGPT: Efficient Spatial Reasoning for Warehouse Environments with 600M ParametersLLM reasoning → Spatial and embodied reasoning +1
OpenViGA: Video Generation for Automotive Driving Scenes by Streamlining and Fine-Tuning Open Source Models with Public DataWorld models → World models for autonomous driving +1
Red Teaming Multimodal Language Models: Evaluating Harm Across Prompt Modalities and ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Evaluating Multimodal Large Language Models on Spoken Sarcasm UnderstandingMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Temporal Reasoning with Large Language Models Augmented by Evolving Knowledge GraphsLLM reasoning → Knowledge- and tool-augmented reasoning
PILOT: Steering Synthetic Data Generation with Psychological & Linguistic Output TargetingSynthetic data and self-improvement for LLMs → LLM synthetic data generation
A Scalable Whole-body Motion Transfer via Implicit Kinodynamic Motion RetargetingOffline RL and imitation learning from data → Real-world robot offline RL
ORCA: An Agentic Reasoning Framework for Hallucination and Adversarial Robustness in Vision-Language ModelsLLM reasoning → Hallucination detection and mitigation +2
SERVAL: Surprisingly Effective Zero-Shot Visual Document Retrieval Powered by Large Vision and Language ModelsMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Analyzing and Mitigating Surface Bias in Code Evaluation MetricsLLM agents and tool use → Software engineering coding agents
Learning in Stackelberg Mean Field Games: A Non-Asymptotic AnalysisDeep reinforcement learning methods → RL for games and combinatorial optimization
Diagnostics of cognitive failures in multi-agent expert systems using dynamic evaluation protocols and subsequent mutation of the processing contextLLM agents and tool use → General LLM agent frameworks and benchmarks
Beyond Spurious Signals: Debiasing Multimodal Large Language Models via Counterfactual Inference and Adaptive Expert RoutingLLM reasoning → Multimodal reasoning +2
MaskAttn-SDXL: Controllable Region-Level Text-To-Image GenerationDiffusion models → Image generation, text-to-image and editing
LowDiff: Efficient Diffusion Sampling with Low-Resolution ConditionDiffusion models → Diffusion foundations and sampling
CoDoL: Conditional Domain Prompt Learning for Out-of-Distribution GeneralizationMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Kuramoto Orientation Diffusion ModelsDiffusion models → Diffusion foundations and sampling
The Distribution Shift Problem in Transportation Networks using Reinforcement Learning and AIDeep reinforcement learning methods → RL for autonomous driving and traffic
Fleming-R1: Toward Expert-Level Medical Reasoning via Reinforcement LearningLLM reasoning → Clinical and biomedical applications
Large Vision Models Can Solve Mental Rotation ProblemsLLM reasoning → Spatial and embodied reasoning
Modeling Transformers as complex networks to analyze learning dynamicsMechanistic interpretability → Knowledge and representations inside LLMs
Autoguided Online Data Curation for Diffusion Model TrainingDiffusion models → Diffusion foundations and sampling
Toxicity Red-Teaming: Benchmarking LLM Safety in Singapore's Low-Resource LanguagesLLM safety and security → Safety alignment, refusal and guardrails
RespoDiff: Dual-Module Bottleneck Transformation for Responsible & Faithful T2I GenerationDiffusion models → Image generation, text-to-image and editing
Walk and Read Less: Improving the Efficiency of Vision-and-Language Navigation via Tuning-Free Multimodal Token PruningEfficient LLM inference and serving → Visual token pruning in multimodal models
Causal Reasoning Elicits Controllable 3D Scene GenerationLLM reasoning → Spatial and embodied reasoning
Calibration-Aware Prompt Learning for Medical Vision-Language ModelsMultimodal LLMs / vision-language models → Medical vision-language models
ScaleCUA: Scaling Open-Source Computer Use Agents with Cross-Platform DataMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Lightweight and Accurate Multi-View Stereo with Confidence-Aware Diffusion ModelDiffusion models → 3D, point cloud and scene generation
Out-of-Sight Embodied Agents: Multimodal Tracking, Sensor Fusion, and Trajectory ForecastingWorld models → World models for autonomous driving
Generalizable Geometric Image Caption SynthesisLLM reasoning → Spatial and embodied reasoning +1
RynnVLA-001: Using Human Demonstrations to Improve Robot ManipulationVision-language-action models → Human-video and latent-action pretraining
FlowRL: Matching Reward Distributions for LLM ReasoningLLM reasoning → Reinforcement learning for reasoning
Beyond Surface Alignment: Rebuilding LLMs Safety Mechanism via Probabilistically Ablating Refusal DirectionLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Fast and Fluent Diffusion Language Models via Convolutional Decoding and Rejective Fine-tuningDiffusion models → Discrete and text diffusion
Unleashing the Potential of Multimodal LLMs for Zero-Shot Spatio-Temporal Video GroundingMultimodal LLMs / vision-language models → Video understanding with MLLMs
Self-Improvement of Language Models by Post-Training on Multi-Agent DebateLLM agents and tool use → Social simulation and multi-agent debate +1
Watermarking and Anomaly Detection in Machine Learning Models for LORA RF FingerprintingLLM safety and security → Watermarking generated content and models
An Evaluation-Centric Paradigm for Scientific Visualization AgentsLLM agents and tool use → Multimodal and embodied agents
AIP: Subverting Retrieval-Augmented Generation via Adversarial Instructional PromptRetrieval-augmented generation → Privacy and security of RAG
Self-Improving Embodied Foundation ModelsSynthetic data and self-improvement for LLMs → Self-evolving LLM agents +1
MedFact-R1: Towards Factual Medical Reasoning via Pseudo-Label AugmentationLLM reasoning → Clinical and biomedical applications +1
ATTS: Asynchronous Test-Time Scaling via Conformal PredictionLLM reasoning → Test-time compute and efficient thinking
From Pixels to Urban Policy-Intelligence: Recovering Legacy Effects of Redlining with a Multimodal LLMMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Taming Video Models for 3D and 4D Generation via Zero-Shot Camera ControlDiffusion models → Video generation
TDRM: Smooth Reward Models with Temporal Difference for LLM RL and InferenceLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Super-Linear: A Lightweight Pretrained Mixture of Linear Experts for Time Series ForecastingMixture-of-experts → MoE for time series and graphs
Digital Twin-based Cooperative Autonomous Driving in Smart Intersections: A Multi-Agent Reinforcement Learning ApproachEnd-to-end autonomous driving → Cooperative and V2X driving
Emergent Alignment via CompetitionLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Real-Time Streaming Mel Vocoding with Generative Flow MatchingDiffusion models → Audio, speech and music
Reinforcement Learning Agent for a 2D Shooter GameOffline RL and imitation learning from data → Imitation learning methods +1
Sample Efficient Experience Replay in Non-stationary EnvironmentsDeep reinforcement learning methods → Deep RL algorithms and representations
AutoEdit: Automatic Hyperparameter Tuning for Image EditingDiffusion models → Image generation, text-to-image and editing +1
No Modality Left Behind: Adapting to Missing Modalities via Knowledge Distillation for Brain Tumor SegmentationKnowledge distillation → Distillation for medical imaging
EchoVLM: Dynamic Mixture-of-Experts Vision-Language Model for Universal Ultrasound IntelligenceMultimodal LLMs / vision-language models → Medical vision-language models +1
Sentinel Agents for Secure and Trustworthy Agentic AI in Multi-Agent SystemsLLM agents and tool use → Agent threats, guardrails and governance
A Novel Task-Driven Diffusion-Based Policy with Affordance Learning for Generalizable Manipulation of Articulated ObjectsOffline RL and imitation learning from data → Robot manipulation imitation learning
Mitigating data replication in text-to-audio generative diffusion models through anti-memorization guidanceDiffusion models → Audio, speech and music
Robot Control Stack: A Lean Ecosystem for Robot Learning at ScaleVision-language-action models → Generalist VLA models and benchmarks
Cross-Modal Knowledge Distillation for Speech Large Language ModelsKnowledge distillation → Speech and audio distillation
Self-Explaining Reinforcement Learning for Mobile Network Resource AllocationDeep reinforcement learning methods → RL for energy, networks and operations
Back to Ear: Perceptually Driven High Fidelity Music ReconstructionDiffusion models → Audio, speech and music
FURINA: Free from Unmergeable Router via LINear Aggregation of mixed expertsMixture-of-experts → Expert routing strategies
Leveraging Reinforcement Learning, Genetic Algorithms and Transformers for background determination in particle physicsDeep reinforcement learning methods → RL for energy, networks and operations
CollabVLA: Self-Reflective Vision-Language-Action Model Dreaming Together with HumanVision-language-action models → Generalist VLA models and benchmarks
MARIC: Multi-Agent Reasoning for Image ClassificationLLM reasoning → Multimodal reasoning +1
Multi-Fidelity Hybrid Reinforcement Learning via Information Gain MaximizationDeep reinforcement learning methods → Deep RL algorithms and representations
[Re] Improving Interpretation Faithfulness for Vision TransformersMechanistic interpretability → Interpreting vision and multimodal models
V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning +1
LLM Agents at the Roundtable: A Multi-Perspective and Dialectical Reasoning Framework for Essay ScoringLLM reasoning → Reasoning benchmarks and evaluation +1
Scalable Multi-Objective Robot Reinforcement Learning through Gradient Conflict ResolutionDeep reinforcement learning methods → RL for games and combinatorial optimization
OnlineMate: An LLM-Based Multi-Agent Companion System for Cognitive Support in Online LearningLLM agents and tool use → Educational agents and math problem solving
MELA-TTS: Joint transformer-diffusion model with representation alignment for speech synthesisDiffusion models → Audio, speech and music
Radiology Report Conditional 3D CT Generation with Multi Encoder Latent diffusion ModelDiffusion models → Medical imaging and segmentation
OpenLens AI: Fully Autonomous Research Agent for Health InfomaticsLLM agents and tool use → Medical and scientific domain agents
Dataset Distillation for Super-Resolution without Class Labels and Pre-trained ModelsDiffusion models → Inverse problems and image restoration
Frame Sampling Strategies Matter: A Benchmark for small vision language modelsMultimodal LLMs / vision-language models → Video understanding with MLLMs
Reasoning over Boundaries: Enhancing Specification Alignment via Test-time DeliberationLLM reasoning → Reasoning for agents, planning and code +1
Designing Latent Safety Filters using Pre-Trained Vision ModelsWorld models → Latent dynamics models and model-based RL
Data Augmentation via Latent Diffusion Models for Detecting Smell-Related Objects in Historical ArtworksDiffusion models → Image generation, text-to-image and editing
Enhancing Retrieval Augmentation via Adversarial CollaborationRetrieval-augmented generation → Privacy and security of RAG
Chain-of-Thought Re-ranking for Image Retrieval TasksLLM reasoning → Multimodal reasoning
On the Use of Agentic Coding: An Empirical Study of Pull Requests on GitHubLLM agents and tool use → Software engineering coding agents
On the Use of Agentic Coding Manifests: An Empirical Study of Claude CodeLLM agents and tool use → Software engineering coding agents
Decoupled Proxy Alignment: Mitigating Language Prior Conflict for Multimodal Alignment in MLLMMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Transcoder-based Circuit Analysis for Interpretable Single-Cell Foundation ModelsMechanistic interpretability → Interpreting biomedical and science models
ToolSample: Dual Dynamic Sampling Methods with Curriculum Learning for RL-based Tool LearningLLM agents and tool use → Reinforcement learning for LLM agents
The Metacognitive Bottleneck: Japanese Riddles Reveal Fundamental Limits of Machine Insight and Self-Evaluation in Reasoning AILLM reasoning → Reasoning benchmarks and evaluation
RationAnomaly: Log Anomaly Detection with Rationality via Chain-of-Thought and Reinforcement LearningLLM reasoning → Reasoning for agents, planning and code
HARNESS: Lightweight Distilled Arabic Speech Foundation ModelsKnowledge distillation → Speech and audio distillation
RealMirror: A Comprehensive, Open-Source Vision-Language-Action Platform for Embodied AIVision-language-action models → Generalist VLA models and benchmarks
LEED: A Highly Efficient and Scalable LLM-Empowered Expert Demonstrations Framework for Multi-Agent Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents
Spatial Audio Motion Understanding and ReasoningLLM reasoning → Audio-language model reasoning
Understanding the Thinking Process of Reasoning Models: A Perspective from Schoenfeld's Episode TheoryLLM reasoning → Mathematical reasoning
Aligning Audio Captions with Human PreferencesLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
MUSE: MCTS-Driven Red Teaming Framework for Enhanced Multi-Turn Dialogue Safety in Large Language ModelsLLM safety and security → Jailbreaking

All 404