Papers

New on arXiv

Mon 27 Jul 2026

399 papers

Pixel Decodability Is Not a Compression Signal: Causally Evaluating Importance Proxies for Visual KV-Cache EvictionMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Lost in Compression: A Controlled Cross-Lingual Audit of Extractive Prompt CompressorsEfficient LLM inference and serving → Prompt and context compression
Evolving Safety Landscape of Multi-modal Large Language Models: A Survey of Emerging Threats and SafeguardsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
When Do Agent Loops Mistake Stagnation for Progress? Self-Evaluation Bias and Externally Grounded Verification in Long-Running Autonomous LLM Agent LoopsLLM agents and tool use → Software engineering coding agents
Agentic AI for Scientific Reasoning in Autonomous Quantum Sensing ExperimentsLLM agents and tool use → Medical and scientific domain agents
How Affect Propagates among LLM Agents: Emergent Emotional Contagion in Crowd SimulationLLM agents and tool use → Social simulation and multi-agent debate
Less Data, Better Alignment: Data-Centric Multi-Evaluator Agreement for Preference OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Interpretable GOHR Agents via Sparse AutoencodersMechanistic interpretability → Steering and interpreting LLM reasoning
LENS: Adaptive Spatio-Temporal Zooming for Keyframe Sampling in Long-Form VideosMultimodal LLMs / vision-language models → Video understanding with MLLMs
Endpoint Replay: Compressing the Recency Buffer in Deep Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
OPERA: Offline Policy-guided Expert Routing and Adaptation for Universal Biomedical Image AnalysisMixture-of-experts → Medical and biomedical MoE
IMPRINT: Image-Conditioned Query Enrichment for Long-Tail Object Goal NavigationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Towards Robust Reinforcement Learning for Small-Scale Language Model AgentsLLM agents and tool use → Reinforcement learning for LLM agents
PLATO: Pointer Learner for Agent and Task OpennessDeep reinforcement learning methods → RL training of language and agent models
Diff-ID: Identity Consistent Facial Image Generation and Morphing via Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
Towards an Agent Operating System - Lessons from Classical and Cloud OSLLM agents and tool use → Software engineering coding agents
DS@GT ARC at CheckThat! 2026: LLM-Based Trace Ranking and Grouped Reward Modeling for Multilingual Numerical Claim VerificationLLM alignment / RLHF / preference optimisation → Process reward models for reasoning
Addressable Recall Compaction for Long Context-Window Control in AI AgentsLLM reasoning → Reasoning for agents, planning and code +1
Lantern: Conflict-Aware Gradient Blending for Physics-Guided Diffusion Models in Calorimeter SimulationDiffusion models → Flow matching, time series and science
SAFAARI: Schema-Aware Framework for Accelerated Advertiser Response IntelligenceLLM agents and tool use → Web, search, and research agents
Authoring Agent Skills: A Software-Engineering ApproachLLM agents and tool use → Agent skill learning and evolution
Chart-Supported or Model-Supplied? Examining MLLM-Generated Claims for Accessible VisualizationMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Calibrated Partial Resets: Preventing Policy Collapse in Continual Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
ALIBI: Adaptive Agentic Attacks on LLM-Based Vulnerability Detectors via Adversarial Code CommentsLLM agents and tool use → Agent threats, guardrails and governance
PerceptionBench: Evaluating Atomic Visual Perception in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation ModelLLM reasoning → Multimodal reasoning +1
Inverse RL Helps Align AI by Imitating HumansLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
TYPO: Instruction-Dense Visual Jailbreaks against Commercial Closed-Source Image-Generation ModelsLLM safety and security → Vision-language and text-to-image safety
Early Detection of Distributed Backdoors in Multi-Agent LLM Systems: A Characterization StudyLLM agents and tool use → Agent threats, guardrails and governance +1
Trusting-Trust Attack against an Entire Linux Distribution through Binary ManipulationLLM safety and security → Backdoor and data poisoning attacks
Beyond "What to Retrieve": Uncertainty in Retrieval-Augmented Code GenerationRetrieval-augmented generation → RAG for code
Agent Retrieval Bench: Evaluating Repository Context Retrieval for Coding AgentsLLM agents and tool use → Software engineering coding agents
Human Preference aligned Tabular SimilarityLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Low-Latency Generative Semantic Communication via Channel-Realization Flow MatchingDiffusion models → Diffusion for compression and wireless
ClinFusion: A Vision-Centric Multimodal LLM System for Holistic Medical UnderstandingMultimodal LLMs / vision-language models → Medical vision-language models
Rethinking Classifier-Free Guidance in On-Policy Diffusion DistillationDiffusion models → Diffusion foundations and sampling
The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic DistillationWorld models → Latent predictive world models +2
ERUnderstand: Evaluating Vision-Language Models on Structured ER DiagramsMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Explainable Reinforcement Learning via Physics-Aware Policy DistillationDeep reinforcement learning methods → Safe and constrained RL
MMOE: Modernizing Diffusion Transformers with Efficient Expert DesignDiffusion models → Discrete and text diffusion
A corrective agentic hybrid RAG and an operations-grounded evaluation for a scientific facilityRetrieval-augmented generation → Agentic RAG, agents and memory +1
Kimi K3: Open Frontier IntelligenceLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Evidence Attribution in Visual Document Understanding without Coordinates or Region LabelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Reason-Mediated Behavioral Models for Auditing LLM Social SimulatorsLLM reasoning → Reasoning benchmarks and evaluation
Sparse Autoencoders Encode Both Concepts and Functions: The Downstream Geometry of Feature EffectsMechanistic interpretability → Sparse autoencoder training and evaluation
APPA: Recoverable Information-Flow Control for Real-World LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
LaRec: Unleashing LLM-based Latent Reasoning for Generative RecommendationLLM reasoning → LLM reasoning for recommendation
Looping Is Not Reliability: State-Bound Evidence and Typed Revision Contracts for Agentic Code RepairLLM agents and tool use → Software engineering coding agents
SIREN: Towards End-to-End Extreme-Weather Early Warning with Experience-Grounded LLM AgentsLLM agents and tool use → Financial trading and forecasting agents
CADER: Confidence-Aware Dynamic Evidence Reasoning for Long-Video UnderstandingLLM reasoning → Multimodal reasoning +1
Evaluating Fuzz Testing for Reinforcement Learning AgentsDeep reinforcement learning methods → RL for autonomous driving and traffic
EgoPlay: Event-Triggered Video Editing for Egocentric StreamsDiffusion models → Video generation
DeCoRAG: Cognitive Decoupling and Semantic-Aware Cropping for Complex Document UnderstandingRetrieval-augmented generation → Multimodal RAG
Task-Conditional Faithfulness Auditing of Multimodal LLMs for Grid DiagnosisMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
FlowCTS: On-policy Continuous Trajectory Supervision of Flow ModelsKnowledge distillation → On-policy and self-distillation for LLMs
DecoupleMix: Decoupled Ratio Search and Convex Allocation for Scalable VLM Data RecipesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
UNIFUSION: Adapting Autoregressive Language Models into Discrete Diffusion under a Unified Reverse-Rate ObjectiveDiffusion models → Discrete and text diffusion
τ: Learning Touch-Augmented Vision-Language-Action Models from Future Visual SupervisionVision-language-action models → Tactile, force-aware and humanoid VLA
What do Reward Models Memorize?LLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
ArmnetBench v0.1: Parallel Real-World Evaluation of Manipulation Policies on a Low-Cost Arm FarmVision-language-action models → Reinforcement learning and post-training
Grounding latent algorithm routing in transformer reasoningLLM reasoning → Prompting and chain-of-thought reasoning
RP-OPSD: Resolution-Privileged On-Policy Self-Distillation for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
Bigger or Cheaper? Scale and Quantization Effects on Uncertainty Signals in Vision-Language Models Under Image DegradationMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
MSVS-VAE: Multi-Scale Anchored VecSet for High-Fidelity 3D ReconstructionDiffusion models → 3D, point cloud and scene generation
DraftExpert: Expansion-Aware Self-Speculative Decoding for End-Device MoE InferenceMixture-of-experts → Efficient MoE training and inference systems +1
Context Is King: How In-Context Specification Shapes the Geometry of ConceptsWorld models → World models for language agents
MAViE: A Multi-scale Adaptive Vision Encoder for Fine-grained Visual Perception and Efficient Multimodal ReasoningLLM reasoning → Multimodal reasoning +1
Mixture-of-Thought-Tokens: Unifying Perception and Reasoning for Free-form Multimodal GroundingLLM reasoning → Multimodal reasoning +1
When LLM Defenses Backfire: Characterizing Safety, Performance, and Cost Trade-offsLLM safety and security → Jailbreaking
MXAttention: Data-Free Optimal Scaling and Pre-Normalization Quantization for MXFP4 AttentionDiffusion models → Video generation
Are Prompt Optimizers Blind? Cross-Modal Visual Feedback for Automatic Prompt OptimizationMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
PRISM: Prompt Refinement via Image-grounded Self-rewarding Mechanism for Text-to-Image GenerationSynthetic data and self-improvement for LLMs → Self-improving MLLMs and synthetic data
Retrieval-Augmented Large Language Models as Components of Cognitive Computing architecture for Regulatory Knowledge ManagementRetrieval-augmented generation → RAG pipeline design and evaluation
Beyond Aggregate Risk: Role-Stratified Conformal Risk Control for LLM Tool CallsLLM agents and tool use → Agent threats, guardrails and governance
Simulating Tenant Responses to Energy Policy Interventions with Transaction-Cost-Aware LLM AgentLLM agents and tool use → Financial trading and forecasting agents
Gubernaut: A Deterministic Homeostatic Controller for Affect-Regulated LLM Agents, Validated Across Independent Model FamiliesLLM agents and tool use → General LLM agent frameworks and benchmarks
Cross-Attention Calibrated Deduplication for Retrieval-Augmented Generation SystemRetrieval-augmented generation → Generation and context integration
DynaCalKV: Key-Value Cache Compression via Head Grouping and Adaptive Rank AllocationEfficient LLM inference and serving → KV cache optimization
Continual-RL for Generalization in Autonomous Racing on the RoboRacer PlatformDeep reinforcement learning methods → Deep RL algorithms and representations
CONSISTRE: A Unified Consistency-Aware Framework for Document-Level Relation Extraction with Large Language ModelsKnowledge distillation → Knowledge distillation for LLMs
Rethinking the Generation Order of Block Diffusion Language ModelsLLM reasoning → Diffusion reasoning and test-time scaling +1
Self-Authored Verification Is Unreliable in Heuristic Self-Improving AgentsSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
PAC-DP: PAC-Bayesian Diffusion Policy LearningDiffusion models → Diffusion and flow policies for robotics
Learning Adaptive Multi-Task Guidance, Navigation, and Control via HypernetworksDeep reinforcement learning methods → RL for robot navigation and UAV control
From Proprietary to Open-Source: Bridging the Distribution Gap via Multi-Agent Protocol Distillation in Agentic SearchLLM agents and tool use → Web, search, and research agents
FeelWorld: Visuo-Tactile World Model for Hierarchical Contact Prediction and PlanningWorld models → Embodied and robot world models
KAP: Bridging the Knowledge Selection-Runtime Consumption Gap in LLM SystemsEfficient LLM inference and serving → Efficient reasoning and agent inference
SILICA: Repurposing Diffusion Priors for Joint Glass Segmentation and Depth EstimationDiffusion models → 3D, point cloud and scene generation
A New Role for Relevance: Guiding Corpus Interaction in Agentic SearchLLM agents and tool use → Web, search, and research agents
Reasoning to Regulate: Chain-of-Thought for Traffic Rule UnderstandingLLM reasoning → Chain-of-thought reasoning for driving
LCMamNet: A Lightweight Cross-scale Mamba Network for Infrared Small Target DetectionState-space sequence models (S4, Mamba) → Point clouds, detection and remote sensing
Just Testing, Move Along: Evasion of LLM-based System Log Interpretation by Prompt InjectionLLM safety and security → Prompt injection and agent security
Falsifiable Commitment Planning for Self-Correcting Web AgentsLLM agents and tool use → Web, search, and research agents
Agent-UCT: Upper Confidence Bounds Applied to Trees for Agentic Workflow Optimization with Cost-AwarenessRetrieval-augmented generation → Agentic RAG, agents and memory +1
A Motion-Aware Vector Quantization Framework with Centroid Reuse for Efficient VLA InferenceVision-language-action models → Efficient VLA inference and adaptation
ViDS: Video Diffusion Shader using 3D Face TrackingDiffusion models → Audio, speech and music
Scaling GUI Agents with Visual State TransitionsLLM agents and tool use → GUI and mobile agents
BeyondFusion: Self-Aligned Latent Diffusion for Calibration-Free Infrared Super-Resolution and Infrared-Visible FusionDiffusion models → Image generation, text-to-image and editing
LU-500: A Logo Benchmark for Concept UnlearningLLM safety and security → Vision-language and text-to-image safety
MemChain: Learning Interpretable Memory Traces for Memory-Augmented LLM AgentsLLM agents and tool use → Agent memory systems

All 399