Papers

New on arXiv

Fri 8 Aug 2025

332 papers

Devstral: Fine-tuning Language Models for Coding Agent ApplicationsLLM agents and tool use → Software engineering coding agents
MAHL: Multi-Agent LLM-Guided Hierarchical Chiplet Design with Adaptive DebuggingLLM agents and tool use → Software engineering coding agents
gpt-oss-120b & gpt-oss-20b Model CardLLM reasoning → Test-time compute and efficient thinking
Latent Fusion Jailbreak: Blending Harmful and Harmless Representations to Elicit Unsafe LLM OutputsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
LLMCARE: early detection of cognitive impairment via transformer models enhanced by LLM-generated synthetic dataSynthetic data and self-improvement for LLMs → Medical synthetic data and agents
SABER: Switchable and Balanced Training for Efficient LLM ReasoningLLM reasoning → Test-time compute and efficient thinking
Learning to Detect Unseen Jailbreak Attacks in Large Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Delta-AttnMask: Attention-Guided Masked Hidden States for Efficient Data Selection and AugmentationMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
MX-AI: Agentic Observability and Control Platform for Open and AI-RANLLM agents and tool use → Agentic AI for wireless networks
Meta-Learning for Speeding Up Large Model Inference in Decentralized EnvironmentsEfficient LLM inference and serving → Serving systems and request scheduling
Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion ForcingDiffusion models → Discrete and text diffusion +1
Multi-Level Safety Continual Projection for Fine-Tuned Large Language Models without RetrainingLLM safety and security → Safety alignment, refusal and guardrails
Learning to control inexact Benders decomposition via reinforcement learningDeep reinforcement learning methods → RL for games and combinatorial optimization
Learning More by Seeing Less: Structure First Learning for Efficient, Transferable, and Human-Aligned VisionKnowledge distillation → General knowledge distillation methods
Role of Large Language Models and Retrieval-Augmented Generation for Accelerating Crystalline Material Discovery: A Systematic ReviewRetrieval-augmented generation → Medical and biomedical RAG
Do Biased Models Have Biased Thoughts?LLM reasoning → Reasoning benchmarks and evaluation
In-Context Reinforcement Learning via Communicative World ModelsWorld models → World models for language agents +1
ClusterMark: Towards Robust Watermarking for Autoregressive Image Generators with Visual Token ClusteringDiffusion models → Robustness, privacy and adversarial aspects +1
Valid Inference with Imperfect Synthetic DataSynthetic data and self-improvement for LLMs → LLM synthetic data generation
Dual-Head Physics-Informed Graph Decision Transformer for Distribution System RestorationOffline RL and imitation learning from data → Decision transformers and sequence modeling +1
CycleDiff: Cycle Diffusion Models for Unpaired Image-to-image TranslationDiffusion models → Image generation, text-to-image and editing
Generalizing Scaling Laws for Dense and Sparse Large Language ModelsMixture-of-experts → MoE large language models
Local Diffusion Models and Phases of Data DistributionsDiffusion models → Diffusion foundations and sampling
BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research AgentLLM agents and tool use → Web, search, and research agents
LLM Unlearning Without an Expert Curated DatasetLLM safety and security → Machine unlearning
Towards Integrated AlignmentLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
CountQA: How Well Do MLLMs Count in the Wild?Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
LightSwitch: Multi-view Relighting with Material-guided DiffusionDiffusion models → 3D, point cloud and scene generation
Effective Training Data Synthesis for Improving MLLM Chart UnderstandingSynthetic data and self-improvement for LLMs → Self-improving MLLMs and synthetic data
HapticLLaMA: A Multimodal Sensory Language Model for Haptic CaptioningLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation ModelsLLM reasoning → Reasoning benchmarks and evaluation +2
LLM Unlearning using Gradient Ratio-Based Influence Estimation and Noise InjectionLLM safety and security → Machine unlearning
ScamAgents: How AI Agents Can Simulate Human-Level Scam CallsLLM reasoning → Reasoning for agents, planning and code +2
SlimInfer: Accelerating Long-Context LLM Inference via Dynamic Token PruningEfficient LLM inference and serving → KV cache optimization
Memp: Exploring Agent Procedural MemoryLLM agents and tool use → Agent memory systems
Shortcut Learning in Generalist Robot Policies: The Role of Dataset Diversity and FragmentationVision-language-action models → Reinforcement learning and post-training
Quantifying Conversation Drift in MCP via Latent PolytopeLLM agents and tool use → Agent threats, guardrails and governance
Sample-efficient LLM Optimization with Reset ReplayLLM reasoning → Reinforcement learning for reasoning +1
V*: An Efficient Motion Planning Algorithm for Autonomous VehiclesEnd-to-end autonomous driving → Imitation-learning end-to-end planners
A Systematic Literature Review of Retrieval-Augmented Generation: Techniques, Metrics, and ChallengesRetrieval-augmented generation → RAG pipeline design and evaluation
FVGen: Accelerating Novel-View Synthesis with Adversarial Video Diffusion DistillationDiffusion models → 3D, point cloud and scene generation
SpeakerLM: End-to-End Versatile Speaker Diarization and Recognition with Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
ActivityDiff: A diffusion model with Positive and Negative Activity Guidance for De Novo Drug DesignDiffusion models → Flow matching, time series and science
Beyond Prompt-Induced Lies: Investigating LLM Deception on Benign PromptsLLM reasoning → Prompting and chain-of-thought reasoning
Harnessing Adaptive Topology Representations for Zero-Shot Graph Question AnsweringMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Unsupervised Partner Design Enables Robust Ad-hoc TeamworkDeep reinforcement learning methods → RL for games and combinatorial optimization
L2Calib: SE(3)-Manifold Reinforcement Learning for Robust Extrinsic Calibration with Degenerate Motion ResilienceDeep reinforcement learning methods → RL for robot navigation and UAV control
M2IO-R1: An Efficient RL-Enhanced Reasoning Framework for Multimodal Retrieval Augmented Multimodal GenerationLLM reasoning → Multimodal reasoning +1
Can Diffusion Models Bridge the Domain Gap in Cardiac MR Imaging?Diffusion models → Medical imaging and segmentation
Towards Balanced Behavior Cloning from Imbalanced DatasetsOffline RL and imitation learning from data → Imitation learning methods
Mixture of Experts Guided by Gaussian Splatters Matters: A new Approach to Weakly-Supervised Video Anomaly DetectionMixture-of-experts → Vision mixture-of-experts
KV Cache Compression for Inference Efficiency in LLMs: A ReviewEfficient LLM inference and serving → KV cache optimization
LLM Robustness Leaderboard v1 --Technical reportLLM safety and security → Jailbreaking
OM2P: Offline Multi-Agent Mean-Flow PolicyOffline RL and imitation learning from data → Offline multi-agent RL
ADPro: a Test-time Adaptive Diffusion Policy via Manifold-constrained Denoising and Task-aware Initialization for Robotic ManipulationDiffusion models → Diffusion and flow policies for robotics
SIFThinker: Spatially-Aware Image Focus for Visual ReasoningLLM reasoning → Multimodal reasoning +1
In-Training Defenses against Emergent Misalignment in Language ModelsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
Membership Inference Attack with Partial FeaturesLLM safety and security → Membership inference attacks
SCAR: State-Space Compression for Scalable AI-Based Network Management of Vehicular ServicesDeep reinforcement learning methods → RL for energy, networks and operations
Towards Unified Image Deblurring using a Mixture-of-Experts DecoderMixture-of-experts → Vision mixture-of-experts
GeoLaux: A Benchmark for Evaluating MLLMs' Geometry Performance on Long-Step Problems Requiring Auxiliary LinesMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
InfoCausalQA:Can Models Perform Non-explicit Causal Reasoning Based on Infographic?LLM reasoning → Multimodal reasoning +1
Reparameterization Proximal Policy OptimizationDeep reinforcement learning methods → RL theory and sample complexity
Affordance-R1: Reinforcement Learning for Generalizable Affordance Reasoning in Multimodal Large Language ModelLLM reasoning → Spatial and embodied reasoning +1
Classification is a RAG problem: A case study on hate speech detectionRetrieval-augmented generation → Privacy and security of RAG
AnomalyMoE: Towards a Language-free Generalist Model for Unified Visual Anomaly DetectionMixture-of-experts → Vision mixture-of-experts
LoRA in LoRA: Towards Parameter-Efficient Architecture Expansion for Continual Visual Instruction TuningMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
SceneJailEval: A Scenario-Adaptive Multi-Dimensional Framework for Jailbreak EvaluationLLM safety and security → Jailbreaking
MA-CBP: A Criminal Behavior Prediction Framework Based on Multi-Agent Asynchronous CollaborationLLM agents and tool use → Multimodal and embodied agents
UR2: Unify RAG and Reasoning through Reinforcement LearningLLM reasoning → Knowledge- and tool-augmented reasoning +1
Fewer Denoising Steps or Cheaper Per-Step Inference: Towards Compute-Optimal Diffusion Model DeploymentDiffusion models → Diffusion foundations and sampling
SLIP: Soft Label Mechanism and Key-Extraction-Guided CoT-based Defense Against Instruction Backdoor in APIsLLM safety and security → Backdoor and data poisoning attacks
Improving Diagnostic Accuracy for Oral Cancer with inpainting Synthesis Lesions Generated Using Diffusion ModelsDiffusion models → Medical imaging and segmentation
Retrieval Augmented Large Language Model System for Comprehensive Drug ContraindicationsRetrieval-augmented generation → Medical and biomedical RAG
SDEval: Safety Dynamic Evaluation for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Less is More: Selective Reflection for Compatible and Efficient Knowledge Distillation in Large Language ModelsKnowledge distillation → Knowledge distillation for LLMs
LLM Serving Optimization with Variable Prefill and Decode LengthsEfficient LLM inference and serving → Serving systems and request scheduling
AURA: Affordance-Understanding and Risk-aware Alignment Technique for Large Language ModelsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
GMF-Drive: Gated Mamba Fusion with Spatial-Aware BEV Representation for End-to-End Autonomous DrivingEnd-to-end autonomous driving → End-to-end system architecture +1
PanelTR: Zero-Shot Table Reasoning Framework Through Multi-Agent Scientific DiscussionLLM reasoning → Knowledge- and tool-augmented reasoning +1
GCHR : Goal-Conditioned Hindsight Regularization for Sample-Efficient Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
You Don't Need Pre-built Graphs for RAG: Retrieval Augmented Generation with Adaptive Reasoning StructuresLLM reasoning → Knowledge- and tool-augmented reasoning +1
UGD-IML: A Unified Generative Diffusion-based Framework for Constrained and Unconstrained Image Manipulation LocalizationDiffusion models → Image generation, text-to-image and editing
Bounding Distributional Shifts in World Modeling through Novelty DetectionWorld models → Latent predictive world models
E-React: Towards Emotionally Controlled Synthesis of Human ReactionsDiffusion models → Audio, speech and music
Q-CLIP: Unleashing the Power of Vision-Language Models for Video Quality Assessment through Unified Cross-Modal AdaptationMultimodal LLMs / vision-language models → Image editing and quality assessment
AdaptInfer: Adaptive Token Pruning for Vision-Language Model Inference with Dynamical Text GuidanceLLM reasoning → Multimodal reasoning +2
ME3-BEV: Mamba-Enhanced Deep Reinforcement Learning for End-to-End Autonomous Driving with BEV-PerceptionEnd-to-end autonomous driving → End-to-end system architecture +2
Fact2Fiction: Targeted Poisoning Attack to Agentic Fact-checking SystemLLM agents and tool use → Web, search, and research agents
RAGTrace: Understanding and Refining Retrieval-Generation Dynamics in Retrieval-Augmented GenerationRetrieval-augmented generation → RAG pipeline design and evaluation
VQAThinker: Exploring Generalizable and Explainable Video Quality Assessment via Reinforcement LearningDeep reinforcement learning methods → RL training of language and agent models
EvolvR: Self-Evolving Pairwise Reasoning for Story Evaluation to Enhance GenerationLLM reasoning → Reasoning benchmarks and evaluation +1
Society of Mind Meets Real-Time Strategy: A Hierarchical Multi-Agent Framework for Strategic ReasoningLLM reasoning → Reasoning for agents, planning and code +1
Fourier Compressor: Frequency-Domain Visual Token Compression for Vision-Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
More Is Better: A MoE-Based Emotion Recognition Framework with Human Preference AlignmentLLM alignment / RLHF / preference optimisation → Medical LLM alignment +1
InstantEdit: Text-Guided Few-Step Image Editing with Piecewise Rectified FlowDiffusion models → Image generation, text-to-image and editing
Learning 3D Texture-Aware Representations for Parsing Diverse Human Clothing and Body PartsDiffusion models → 3D, point cloud and scene generation
Temporal Self-Rewarding Language Models: Decoupling Chosen-Rejected via Past-FutureLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
EC2MoE: Adaptive End-Cloud Pipeline Collaboration Enabling Scalable Mixture-of-Experts InferenceMixture-of-experts → Efficient MoE training and inference systems
Improved Sub-Visible Particle Classification in Flow Imaging Microscopy via Generative AI-Based Image SynthesisDiffusion models → Flow matching, time series and science

All 332