Papers

New on arXiv

Wed 24 Jun 2026

426 papers

Skim and Skip: Hierarchical Adaptive Inference for Efficient Multimodal RetrievalMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
I-CARE: Analysis of interference-related phenomena in a controllable, diverse and representative unlearning setting for text-to-image modelsLLM safety and security → Vision-language and text-to-image safety
Agent Seer: Synthesizing Scenarios from Specification UnderstandingLLM agents and tool use → Tool learning and function calling
TH-GNN: Heterogeneous Temporal Graph Neural Networks for LLM-Agent Shilling Attack DetectionLLM agents and tool use → Agent threats, guardrails and governance
GRAFT: Adaptive DLM-Based Draft Tree Construction with Target-Distilled Edge ScoringEfficient LLM inference and serving → Speculative decoding
Optimization and Constraint Modeling using LLMs with a Retrieval Augmented Generation ProcessRetrieval-augmented generation → RAG pipeline design and evaluation
What Transfers from Text to Vision? Capability Scaling Laws and Transfer Dynamics for VLMsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
SpecPrefetch: Parameter-Efficient Expert Prefetching for Sparse MoE Foundation ModelsMixture-of-experts → Efficient MoE training and inference systems
DocHRL: A Hierarchical Reinforcement Learning Framework for Cost-Optimised Document ClassificationDeep reinforcement learning methods → RL training of language and agent models
Reason Before You Retrieve: Agentic Planning for Multi-modal RAGRetrieval-augmented generation → Multimodal RAG +1
CRAFT: Learn the Schema, Execute the PlanLLM agents and tool use → Software engineering coding agents
FinBench: Time-Gated Calibration and Uncertainty Benchmarking for Agentic Financial ForecastingLLM agents and tool use → Financial trading and forecasting agents
Closed-Loop Control with Rule-Aligned Small Language Models and Multi-Agent Self-CorrectionLLM agents and tool use → General LLM agent frameworks and benchmarks
Reinforcement Learning for Software Vulnerability Analysis: A Systematic Review with Emphasis on C/C++ Source Code and Static AnalysisDeep reinforcement learning methods → RL training of language and agent models
Vision-driven Preference Synthesis for Mitigating Hallucinations in VLMsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs +1
CLOSER-VLN: Closed-Loop Self-Verified Retrieval-Augmented Reasoning for Aerial Vision-Language NavigationRetrieval-augmented generation → Multimodal RAG
SidConArena: An Environment Evaluating Agents in Open-Ended,Positive-Sum Bargaining GameLLM agents and tool use → Social simulation and multi-agent debate
MKG-RAG-Bench: Benchmarking Retrieval in Multimodal Knowledge Graph-Augmented GenerationRetrieval-augmented generation → Multimodal RAG
Data-driven Machine Learning Cannot Reach Symbolic-level Logical Reasoning -- The Limit of the Scaling LawLLM reasoning → Reasoning benchmarks and evaluation
ProvenAI: Provenance-Native Traces of Evidence in Generated AnswersRetrieval-augmented generation → RAG robustness and adaptive retrieval
Exploring the Intrinsic Geometry of Diffusion Models with Constrained Inverse KinematicsDiffusion models → Diffusion foundations and sampling
ProfileFoundry: A Synthetic Person-Object Substrate for Privacy, Memory, and Tool-Use Evaluation in LLM AgentLLM agents and tool use → Agent threats, guardrails and governance
Deterministic Pareto-Optimal Policy Synthesis for Multi-Objective Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
At the Edge of Understanding: Sparse Autoencoders Trace The Limits of Transformer GeneralizationMechanistic interpretability → Circuits and mechanistic analysis
MPC-Injection: Biasing Off-Policy Locomotion RL Toward Controller-Induced Behavior BasinsDeep reinforcement learning methods → RL for robot locomotion and manipulation
Sampling sea state using a diffusion modelDiffusion models → Flow matching, time series and science
Staying VIGILant: Mitigating Visual Laziness via Counterfactual Visual Alignment in MLLMsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
What Do Deepfake Benchmarks Measure? An Audit Using Frozen Self-Supervised RepresentationsMechanistic interpretability → Probing deception, bias and beliefs
Verifying Intent and Harm: A Unified Defense Against LLM-Generated ThreatsLLM safety and security → Prompt injection and agent security
Narration-of-Thought: Inference-Time Scaffolding for Defeasible Ethical Reasoning in Large Language ModelsLLM reasoning → Reasoning benchmarks and evaluation
Instruction Bleed: Cross-Module Interference in Prompt-Composed Agentic SystemsLLM agents and tool use → Agent threats, guardrails and governance
OpenFinGym: A Verifiable Multi-Task Gym Environment for Evaluating Quant AgentsLLM agents and tool use → Financial trading and forecasting agents
What We are Missing in Multimodal LLM Evaluation?Multimodal LLMs / vision-language models → MLLM training and visual instruction tuning
How Do Tool-Augmented LLM Agents Perform on Real-World Energy Analytics Tasks?LLM agents and tool use → Financial trading and forecasting agents
Feasibility-Aware Security-Constrained Unit Commitment via Hybrid Soft Actor-Critic with Quantum-Sampled FeaturesDeep reinforcement learning methods → Quantum reinforcement learning
Mesh-RL: Coupled subgrid reinforcement learningDeep reinforcement learning methods → RL theory and sample complexity
EVOM: Agentic Meta-Evolution of Actor-Critic Architectures for Reinforcement LearningLLM agents and tool use → General LLM agent frameworks and benchmarks
KRVF: A Source-Aware Semantic Voxel World Representation for Edge Mobile ManipulationWorld models → Embodied and robot world models
Racing a Wheeled Quadruped: Active Load Transfer Mitigation via Model Predictive ControlDeep reinforcement learning methods → RL for autonomous driving and traffic
The Verification Horizon: No Silver Bullet for Coding Agent RewardsLLM agents and tool use → Software engineering coding agents
The Red Queen G"odel Machine: Co-Evolving Agents and Their EvaluatorsLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Augmentation with Dilution: A Large-Scale Empirical Study of Human Contributor Ecosystems After AI Coding Agent AdoptionLLM agents and tool use → Software engineering coding agents
GeMoE: Gating Entropy is All You Need for Uncertainty-aware Adaptive Routing in MoE-based Large Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
TEMPO-Diffusion: Temporally Exposed Malicious Poisoning of Diffusion ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
Fast LeWorldModelWorld models → Latent predictive world models
CyberChainBench: Can AI Agents Secure Smart Contracts Against Real-World On-Chain Vulnerabilities?LLM agents and tool use → Agent threats, guardrails and governance
Knowledge-augmented Agentic AI for Mental Health Medication Information SeekingLLM agents and tool use → Medical and scientific domain agents
Agentic Analysis for Agentic Infrastructure: An LLM-Powered Pipeline for Comparative Governance of DAO and Corporate AI ProtocolsLLM agents and tool use → General LLM agent frameworks and benchmarks
From Structure to Synergy: A Survey of Vision-Language Perception Paradigm Evolution in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Tracking the Turn: Mamba-Powered Human Orientation Detection using UWBState-space sequence models (S4, Mamba) → Video, motion and generative Mamba models
LiMoDE: Rethinking Lifelong Robot Manipulation from a Mixture-of-Dynamic-Experts PerspectiveMixture-of-experts → MoE in reinforcement learning and robotics
Implementation of reinforcement learning in chemical reaction networks: application to phototaxis as curiosity-driven explorationDeep reinforcement learning methods → RL for robot navigation and UAV control
Learning Action Priors for Cross-embodiment Robot ManipulationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
On-Policy Self-Distillation with Sampled Demonstrations Reduces Output DiversityKnowledge distillation → On-policy and self-distillation for LLMs
MVTrack4Gen: Multi-View Point Tracking as Geometric Supervision for 4D Video GenerationDiffusion models → Video generation
Neglected Free Lunch from Post-training: Progress Advantage for LLM AgentsLLM reasoning → Reinforcement learning for reasoning +2
Same Evidence, Different Answer: Auditing Order Sensitivity in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Deep Reinforcement Learning-Enhanced Event-Triggered Data-Driven Predictive Control for a 3D Cable-Driven Soft Robotic ArmDeep reinforcement learning methods → RL for robot locomotion and manipulation
How Robust is OCR-Reasoning? Evaluating OCR-Reasoning Robustness of Vision-Language Models under Visual PerturbationsLLM reasoning → Multimodal reasoning +1
Detect, Unlearn, Restore: Defending Text Summarization Models Against Data PoisoningLLM safety and security → Backdoor and data poisoning attacks
TriViewBench: Controlled Complexity Scaling for Multi-View Structural Reasoning in MLLMsLLM reasoning → Multimodal reasoning +1
Why Multi-Step Tool-Use Reinforcement Learning Collapses and How Supervisory Signals Fix ItLLM agents and tool use → Reinforcement learning for LLM agents
In-Context World Modeling for Robotic ControlWorld models → Embodied and robot world models +1
Privacy Vulnerabilities of Attention Layers in Tabular Foundation Models and Protection of High-Risk QueriesLLM safety and security → Membership inference attacks
The Tatoxa System for Text Detoxification in Low-Resource Languages: The Case of TatarLLM safety and security → Safety alignment, refusal and guardrails
FORCE: Efficient VLA Reinforcement Fine-Tuning via Value-Calibrated Warm-up and Self-DistillationVision-language-action models → Reinforcement learning and post-training +1
Hierarchical Reinforcement Learning for Neural Network Compression (HiReLC): Pruning and QuantizationDeep reinforcement learning methods → Deep RL algorithms and representations
Autodata: An agentic data scientist to create high quality synthetic dataLLM agents and tool use → Web, search, and research agents
Weave of Formal ThoughtLLM reasoning → Test-time compute and efficient thinking
Action ControlNet: A Lightweight Delay-Aware Adapter for Smooth Asynchronous Control in Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation
InvestPhilBench: A Multi-Layer Benchmark for Evaluating Large Language Model Procedural Reasoning in Expert Investment PhilosophyLLM reasoning → Reasoning for agents, planning and code
Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight VectorsMixture-of-experts → MoE theory and gating functions
Agentic System as Compressor: Quantifying System Intelligence in BitsLLM agents and tool use → Agent memory systems
Knowledge Cascade: Reverse Knowledge Distillation on Nonparametric Multivariate Functional EstimationKnowledge distillation → General knowledge distillation methods
Improving Richardson--Lucy Deconvolution with Diffusion Priors for Fluorescence MicroscopyDiffusion models → Inverse problems and image restoration
In-context Region-based Drag: Drag Any Region to Any ShapeDiffusion models → Image generation, text-to-image and editing
Manipulation Is Task-Dependent: A Multi-Axis, Multi-Environment Evaluation of Frontier LLMsLLM agents and tool use → Agent skill learning and evolution
Enhancing Brain MRI Anomaly Detection and Reasoning with ROI Rethink and Synthetic DataLLM reasoning → Clinical and biomedical applications +2
Semantic Consistency Policy Optimization for Reinforcement Learning of LLM AgentsLLM agents and tool use → Reinforcement learning for LLM agents
Graph it first! Enabling Reasoning on Long-form Egocentric Videos through Scene GraphsLLM reasoning → Spatial and embodied reasoning +1
Edges Before Embeddings: A Confidence-Aware Blur Gate for Vision-Language PipelinesMultimodal LLMs / vision-language models → Visual token pruning and efficiency
AI Snitches Get Glitches: Towards Evading Agentic SurveillanceLLM agents and tool use → Agent threats, guardrails and governance +1
MiniOpt: Reasoning to Model and Solve General Optimization Problems with Limited ResourcesLLM reasoning → Reinforcement learning for reasoning
SARA: Unlocking Multilingual Knowledge in Mixture-of-Experts via Semantically Anchored Routing AlignmentMixture-of-experts → Expert routing strategies
Beyond Function Calling: Benchmarking Tool-Using Agents under Tool-Environment UnreliabilityLLM agents and tool use → Tool learning and function calling
ROAD-VLA: Robust Online Adaptation via Self-Distillation for Vision-Language-Action ModelsVision-language-action models → Reinforcement learning and post-training +1
Do Encoders Suffice? A Systematic Comparison of Encoder and Decoder Safety Judges for LLM Adversarial EvaluationLLM safety and security → Safety alignment, refusal and guardrails
Uncertainty Quantification for Computer-Use Agents: A Benchmark across Vision-Language Models and GUI Grounding DatasetsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs +1
Dual Distribution Estimation for Zero-shot Noisy Test-Time Adaptation with VLMsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
OPERA: Aligning Open-Ended Reasoning via Objective Perplexity-based Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning +1
RAS: Measuring LLM Safety Through Refusal AlignmentLLM safety and security → Safety alignment, refusal and guardrails
UniTeD: Unified Temporal Diffusion for Joint Perception and Planning in Autonomous DrivingDiffusion models → Diffusion and flow policies for robotics +1
Tracing Target Answers in Poisoned Retrieval Corpora via Token Influence AttributionRetrieval-augmented generation → Privacy and security of RAG
GUI agent: Guided Exploration of User-Sensitive ScreensLLM agents and tool use → GUI and mobile agents
Falcon: Functional Assembly and Language for Compositional Reasoning in X-rayLLM reasoning → Clinical and biomedical applications +1
Memory-Efficient Policy Libraries with Low-Rank Adaptation in Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Average-Power-Budgeted Underwater Vehicle Control via Constrained Reinforcement LearningDeep reinforcement learning methods → RL for robot navigation and UAV control
Joint Residual Reweighting for Classifier Free Guidance in Flow-Matching Zero-Shot TTSDiffusion models → Audio, speech and music
Learning Subset-Shared Invariances for Domain Generalization with Mixture-of-ExpertsMixture-of-experts → Expert routing strategies
Towards a Dynamic and Fixed-budget Memory Bank for Efficient Streaming Video UnderstandingMultimodal LLMs / vision-language models → Video understanding with MLLMs

All 426