Papers

New on arXiv

Tue 23 Jun 2026

477 papers

Agents Don't Paginate: First-Chunk Selection for LLM Tool ResponsesLLM agents and tool use → Tool learning and function calling
VA-DPO: Valence-Arousal Direct Preference Optimization for Controllable Emotion Generation in Language ModelsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Solving Is Not Drawing: A Benchmark for Diagrammatic Reasoning in Olympiad GeometryLLM reasoning → Spatial and embodied reasoning
When Chain-of-Thought Helps and When It Hurts: An Empirical Investigation of the Serial-Depth Bottleneck in LLM ReasoningLLM reasoning → Prompting and chain-of-thought reasoning
Governing Execution Risk in Agentic AI Systems: A Trajectory-Guided Framework for Red TeamingLLM agents and tool use → Agent threats, guardrails and governance +1
LLM Unlearning for Cyber Defense: A Survey on Methods, Challenges, and Emerging ThreatsLLM safety and security → Prompt injection and agent security
EvoClawBench: Can Agents Learn Reusable Skills from Their Own Runs?LLM agents and tool use → Agent skill learning and evolution
The Verifier is the Curriculum: Precision Sets the Return on Search in Code Self-DistillationKnowledge distillation → On-policy and self-distillation for LLMs
Detecting and Controlling Sycophancy with Cascading Linear FeaturesMechanistic interpretability → Activation steering methods
Spatio-Temporal Retrieval-based Priors for Adaptive Computational Teaching in DrivingEnd-to-end autonomous driving → Imitation-learning end-to-end planners
Reflective VLA: In-Context Action Consequences Make VLAs GeneralizeVision-language-action models → Efficient VLA inference and adaptation
ASAP: Agent-System Co-Design for Wall-Clock-Centered Auto HPO Research for ML ExperimentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Heuresis: Search Strategies for Autonomous AI Research Agents Across Quality, Diversity and NoveltyLLM agents and tool use → Web, search, and research agents
SoK: AI Secure Code Generation: Progress, Pitfalls, and Paths ForwardLLM agents and tool use → Agent threats, guardrails and governance
To Isolate or to Score? Model-Adaptive Assessment for Cost-Efficient Multi-Agent RAGRetrieval-augmented generation → Agentic RAG, agents and memory
What Intermediate Layers Know: Detecting Jailbreaks from Entropy DynamicsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Learning Perceptive Platform Adaptive Locomotion Controllers for Quadrupedal RobotsDeep reinforcement learning methods → RL for robot locomotion and manipulation
TRUSTMEM: Learning Trustworthy Memory Consolidation for LLM Agents with Long-Term MemoryLLM agents and tool use → Agent memory systems
Toward Low-Latency Vision-Language Models with Doubly-Correct Predictions in Egocentric Visual UnderstandingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Reward-Conditioned Attention: How Reward Design Shapes What Autonomous Driving Agents SeeEnd-to-end autonomous driving → Reinforcement learning for driving +2
Forget to Improve: On-Device LLM-Agent Continual Learning via Budget-Curated MemoryLLM agents and tool use → Agent threats, guardrails and governance
Speculative Decoding at Temperature Zero: A Scoped Safety-Invariance Screen with a 48,072-Sample ExpansionEfficient LLM inference and serving → Speculative decoding
How Modular Is a Frontier Mixture-of-Experts? A Pre-registered Causal Test in Which Apparent Expert Modularity Mostly DissolvesMixture-of-experts → MoE theory and gating functions
Speculation at a Distance: Where Edge-Cloud Speculative Decoding Actually Pays OffEfficient LLM inference and serving → Early exit and edge inference
Are We There Yet? Exploring the Capabilities of MLLMs in Assistive AI ApplicationsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Do vision-language models search like humans? Reasoning tokens as a reaction-time analog in classic visual-search paradigmsLLM reasoning → Multimodal reasoning +1
LLM-Based Scientific Peer Review: Methods, Benchmarks, and Reliability ChallengesLLM safety and security → Prompt injection and agent security
Chorus II: Cross-Request Sparsity Reuse for Efficient Image-to-Video GenerationDiffusion models → Video generation
Do Thinking Tokens Help with Safety?LLM reasoning → Reasoning for agents, planning and code
Bias-Controlled Primal-Dual Natural Actor-Critic: Optimal Rates for Constrained Multi-Objective Average-Reward RLDeep reinforcement learning methods → RL for games and combinatorial optimization
Erased, but Not Gone: Output Forgetting Is Not True ForgettingLLM safety and security → Machine unlearning
Geo-Strat-RL: Learning Geological Event Reasoning from Verifiable TasksLLM reasoning → Spatial and embodied reasoning
A Zeroth-Order Deep Learning Method for Fully Nonlinear Parabolic Partial Differential Equations with Unknown CoefficientsDeep reinforcement learning methods → RL theory and sample complexity
ExTra: Exploratory Trajectory Optimization for Language Model Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning
The Geometry of Sequential Learning: Lie-Bracket Prediction of Transfer OrderLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Diagnosing and Mitigating Compounding Failures in Agentic Persuasion via Taxonomic Strategy RetrievalLLM agents and tool use → Social simulation and multi-agent debate
Project Auto-World: Towards Automated Benchmarking of Neural Relational ReasonersLLM reasoning → Reasoning benchmarks and evaluation
Evidence for feature-specific error correction in LLMsMechanistic interpretability → SAE features and their applications
Curvature-Guided Mixing for MLLM AdaptationMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Towards Scalable Multi-Task Reinforcement Learning with Large Decision ModelsDeep reinforcement learning methods → Deep RL algorithms and representations
Dustin: Draft-Augmented Sparse Verification for Efficient Long-Context Generation with Speculative DecodingEfficient LLM inference and serving → Speculative decoding
Perfect Detection, Failed Control: The Geometry of Knowing vs. Steering in Language ModelsMechanistic interpretability → Probing deception, bias and beliefs
MacroLens: A Multi-Task Benchmark for Contextual Financial Reasoning under Macroeconomic ScenariosLLM reasoning → Reasoning for agents, planning and code
Supervised Reinforcement Learning for the Coordination of Distributed Energy ResourcesDeep reinforcement learning methods → RL for energy, networks and operations
Conformal Orbit-Valid Trust Horizons for Equivariant World ModelsWorld models → Latent predictive world models
When Do Conservation Laws Survive Learned Representations? Certified Horizons for Latent World ModelsWorld models → Latent predictive world models
DiffusionBench: On Holistic Evaluation of Diffusion TransformersDiffusion models → Diffusion foundations and sampling
InSight: Self-Guided Skill Acquisition via Steerable VLAsVision-language-action models → Generalist VLA models and benchmarks +1
FLAT: Feedforward Latent Triangle Splatting for Geometrically Accurate Scene GenerationDiffusion models → 3D, point cloud and scene generation
FLUX3D: High-Fidelity 3D Gaussian Generation with Diffusion-Aligned Sparse RepresentationDiffusion models → 3D, point cloud and scene generation
"Zooming In" on Agentic Web Browsers as Assistive Technologies: A Case Study with a Low-Vision Technology ExpertLLM agents and tool use → GUI and mobile agents
OpenThoughts-Agent: Data Recipes for Agentic ModelsLLM reasoning → Reasoning for agents, planning and code +1
IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image GenerationLLM reasoning → Multimodal reasoning +1
World Models in Pieces: Structural Certification for General AgentsWorld models → World models for language agents
Grading the Grader: Lessons from Evaluating an Agentic Data Analysis SystemLLM agents and tool use → General LLM agent frameworks and benchmarks
Solving Inverse Problems of Chaotic Systems with Bidirectional Conditional Flow MatchingDiffusion models → Flow matching, time series and science
SHERLOC: Structured Diagnostic Localization for Code Repair AgentsLLM agents and tool use → Software engineering coding agents
High-Fidelity Synthetic Transmission Electron Microscopy Image Generation Using Diffusion Probabilistic Models for Data-Limited Semiconductor MetrologyDiffusion models → Flow matching, time series and science
MANGO: Automated Multi-Agent Test Oracle Generation for Vision-Language-Action ModelsVision-language-action models → Failure recovery and robustness
Vision-Language Model Reasoning for Contextual Semantic Mapping in IntralogisticsLLM reasoning → Spatial and embodied reasoning +1
Paying to Know: Micro-Transaction Markets for Verified Product Information in Agentic E-CommerceLLM agents and tool use → Social simulation and multi-agent debate
Are We Ready For An Agent-Native Memory System?Retrieval-augmented generation → Agentic RAG, agents and memory +1
UniDrive: A Unified Vision-Language and Grounding Framework for Interpretable Risk Understanding in Autonomous DrivingLLM reasoning → Chain-of-thought reasoning for driving +2
Beyond U-Net: A Latent-Representation-Aligned Skip-Free Backbone for Flow-Matching Speech EnhancementDiffusion models → Audio, speech and music
World Value Models for Robotic ManipulationWorld models → Embodied and robot world models
BioMedVR: Confusion-Aware Mixture-of-Prompt Experts for Biomedical Visual ReprogrammingMultimodal LLMs / vision-language models → Medical vision-language models
SER: Learning to Ground Video Reasoning with Semantic Evidence RewardsLLM reasoning → Multimodal reasoning
Evaluating the Interpretability of Sparse Autoencoders with Concept AnnotationsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning +1
SupplyNet: Supporting Visual Exploratory Learning in Supply Chain via Contextual Multi-Agent SimulationLLM agents and tool use → Social simulation and multi-agent debate
Automated Summarization of Software Documents: An LLM-based Multi-Agent ApproachLLM agents and tool use → Software engineering coding agents
LaGO: Latent Action Guidance for Online Reinforcement LearningDeep reinforcement learning methods → RL training of language and agent models
Agentic Collaborative Cognition for Zero-Shot 3D UnderstandingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
CineCap: Structured Reasoning with Spatio-Temporal Anchors for Cinematographic Video CaptioningLLM reasoning → Multimodal reasoning
Beyond Monotonic Progress: Retry-Supervised Value Learning for Robot ImitationOffline RL and imitation learning from data → Imitation learning methods
SAFARI: Scaling Long Horizon Agentic Fault Attribution via Active InvestigationLLM agents and tool use → Software engineering coding agents
Privacy-Preserving RAG via Multi-Agent Semantic Rewriting: Achieving Confidentiality Without Compromising Contextual FidelityRetrieval-augmented generation → Privacy and security of RAG +1
Themis: An explainable AI-enabled framework for Reinforcement Learning with Human FeedbackLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
ScaleToT: Generalizing Structured LLM Reasoning for Billion-Scale Low-Activity User ModelingLLM reasoning → Test-time compute and efficient thinking
Qwen-AgentWorld: Language World Models for General AgentsWorld models → World models for language agents +1
MemAudit: Auditing Long-Term Agent Memory via Hidden User-State RecoveryLLM agents and tool use → Agent memory systems
AdversaBench: Automated LLM Red-Teaming with Multi-Judge Confirmation and Cross-Model TransferabilityLLM safety and security → Prompt injection and agent security
Heterogeneous Knowledge Distillation via Geometry Decoupling and Momentum-Aware Gradient RegulationKnowledge distillation → General knowledge distillation methods
Are Text-to-Image Models Inductivist Turkeys? A Counterfactual Benchmark for Causal ReasoningLLM reasoning → Multimodal reasoning
Reasoning as Attractor Dynamics: Latent Memory Retrieval via Gibbs-Weighted Energy MinimizationLLM reasoning → Test-time compute and efficient thinking
PointVG-R: Internalizing Geometric Reasoning in MLLMs for Precise Pointing Localization via Visual Chain of ThoughtLLM reasoning → Spatial and embodied reasoning
ForensicsTok: Forensics-Guided Tokenized Modeling for Image Tampering LocalizationMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Governed Shared Memory for Multi-Agent LLM SystemsLLM agents and tool use → Agent memory systems
NatureBench: Can Coding Agents Match the Published SOTA of Nature-Family Papers?LLM agents and tool use → Web, search, and research agents
AGORA: An Archive-Grounded Benchmark for Agentic Workplace Document ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning +1
VisCritic: Visual State Comparison as Process Reward for GUI AgentsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +2
What Do Flow-Based Inverse Solvers Approximate? A Posterior-Transport ViewDiffusion models → Flow matching, time series and science
Reinforcement Learning for Computer-Use Agents with Autonomous EvaluationLLM agents and tool use → GUI and mobile agents +1
A specialized reasoning large language model for accelerating rare disease diagnosis: a randomized AI physician assistance trialLLM reasoning → Clinical and biomedical applications
CrossPool: Efficient Multi-LLM Serving for Cold MoE Models through KV-Cache and Weight DisaggregationMixture-of-experts → Efficient MoE training and inference systems +1
Red-Teaming the Agentic Red-TeamLLM agents and tool use → Agent threats, guardrails and governance
MambaRaw: Selective State Space Modeling for Efficient 4K Raw Image ReconstructionState-space sequence models (S4, Mamba) → Mamba for image restoration
G3VLA: Geometric inductive bias for Vision-Language-Action ModelsVision-language-action models → 3D and spatial-aware VLA
The Latent Bridge: A Continuous Slow-Fast Channel for Real-Time Game AgentsLLM agents and tool use → Reinforcement learning for LLM agents
CompressKV: Semantic-Retrieval-Guided KV-Cache Compression for Resource-Efficient Long-Context LLM InferenceEfficient LLM inference and serving → KV cache optimization
Bayesian control for coding agentsLLM agents and tool use → General LLM agent frameworks and benchmarks

All 477