Papers

New on arXiv

Thu 9 Jul 2026

344 papers

When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability RoutingLLM reasoning → Time series reasoning +1
QuantumNovelty: A Skill-Orchestrating Language Agent for Referee-Style Review and Patentability Screening of Quantum Papers and PatentsLLM agents and tool use → Web, search, and research agents
Role Steering of Language Models for Social SimulationsLLM agents and tool use → Social simulation and multi-agent debate
Do Active SAE Feature Planes Carry More Holonomy? A Preregistered Reversal in GemmaMechanistic interpretability → Sparse autoencoder training and evaluation
Trivial Prompt Reframing Bypasses Safety Guardrails in Google's MedGemma-4BLLM safety and security → Safety alignment, refusal and guardrails
Metadata-Free Meta-Reweighted Direct Preference Optimization under Noisy Preference LabelsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Agentic Context Learning with Self-Discovered SpecificationLLM agents and tool use → Agent memory systems
StreamDQ: Near-Memory Weight DeQuantization in Custom HBM for Scalable AI Inference AccelerationEfficient LLM inference and serving → KV cache quantization
SCATE: Learning to Supervise Coding Agents for Cost-Effective Test GenerationLLM agents and tool use → Software engineering coding agents
CLAP: Direct VLM-to-VLA Adaptation via Language-Action GroundingVision-language-action models → Generalist VLA models and benchmarks
Eluna: An Agentic LLM System for Automating Warehouse Operations with Reasoning and Task ExecutionLLM reasoning → Reasoning for agents, planning and code +1
TSRouter: Dynamic Modality-Model Selection for Time Series ReasoningLLM reasoning → Time series reasoning
Better Harnesses, Smaller Models: Building 90% Cheaper Agents via Automated Harness AdaptationLLM agents and tool use → Software engineering coding agents
BlockServe: Block-Grained Continuous Batching for High-Throughput Diffusion LLM ServingEfficient LLM inference and serving → Diffusion LLM and MoE inference
SafeExplorer: An Unbiased Policy Gradient for Reinforcement Learning with Recovery InterventionsDeep reinforcement learning methods → Safe and constrained RL
GATS: Graph-Augmented Tree Search with Layered World Models for Efficient Agent PlanningWorld models → World models for language agents +1
Optimizing Against Safety Representations: Activation-Guided Adversarial Suffixes and the Geometry of RefusalLLM safety and security → Safety alignment, refusal and guardrails
Decoupled Illumination Priors for Spatially Controllable Multi-View Indoor Scene RelightingDiffusion models → 3D, point cloud and scene generation
FlowDAgger: Human-in-the-Loop Adaptation of Generative Robot Policies in Latent SpaceDiffusion models → Diffusion and flow policies for robotics
How are linear representations learned? Exact solutions to the dynamics of abstractionMechanistic interpretability → Circuits and mechanistic analysis
Mixture of Probes: Learning from Privileged Modalities in Multimodal LLMs Through ProbingMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
LongE2V: Long-Horizon Event-based Video Reconstruction, Prediction, and Frame Interpolation with Video Diffusion ModelsDiffusion models → Video generation
UniClawBench: A Universal Benchmark for Proactive Agents on Real-World TasksMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video GeneratorsDiffusion models → Video generation +1
OpenCoF: Learning to Reason Through Video GenerationLLM reasoning → Multimodal reasoning
Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea GenerationLLM reasoning → Knowledge- and tool-augmented reasoning
Score Accuracy Along the Forward Diffusion Does Not Certify Numerical Stability in Diffusion SamplingDiffusion models → Diffusion foundations and sampling
AUTOPILOT VQA: Benchmarking Vision-Language Models for Incident-Centric Dashcam UnderstandingMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection
Latent Memory Palace: Reasoning for Control as Autoregressive Variational InferenceLLM reasoning → Test-time compute and efficient thinking
HumanForge: A Human-Centric Deepfake Video Benchmark with Multi-Agent Forgery RationalesDiffusion models → Image generation, text-to-image and editing
MPFlow: Learning Budgeted Max-Flow Optimization on the Lightning Network with Deep Graph Reinforcement LearningDeep reinforcement learning methods → Adversarial robustness and RL in finance
Do You Need a Frontier Model as a Citation Verifier? Benchmarking Rubric LLMs for Deep-Research Source AttributionLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
A Practical Investigation of Training-free Relaxed Speculative DecodingEfficient LLM inference and serving → Speculative decoding
SolarChain-Eval: A Physics-Constrained Benchmark for Trustworthy Economic Agents in Decentralized Energy MarketsLLM agents and tool use → Financial trading and forecasting agents
WebSwarm: Recursive Multi-Agent Orchestration for Deep-and-Wide Web SearchLLM agents and tool use → Web, search, and research agents
Formal Mechanisms for Market Stability in Self-Interested Agent Societies: A Marketplace Simulation StudyLLM agents and tool use → Financial trading and forecasting agents
Multi-Modal, Multi-Environment Machine Teaching for Robust Reward LearningLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
DominoTree: Conditional Tree-Structured Drafting with Domino for Speculative DecodingEfficient LLM inference and serving → Speculative decoding
When Structured Sparse Autoencoders Learn Consistent Concepts Across ModalitiesMechanistic interpretability → Interpreting vision and multimodal models
It Takes a MAESTRO To Prune Bad ExpertsMixture-of-experts → Expert routing strategies
FabriVLA: A Lightweight Vision-Language-Action Model with Conformal Action Chunk UncertaintyVision-language-action models → Efficient VLA inference and adaptation
Switch-Reasoner: Learn When to Think in Multitask Mixtures via Reinforcement LearningLLM reasoning → Test-time compute and efficient thinking +2
SMetric: Rethink LLM Scheduling for Serving Agents with Balanced Session-centric SchedulingLLM agents and tool use → Software engineering coding agents
Cross-seed explainability using Procrustes-conditioned Joint End-to-end Top-K Sparse AutoencodersMechanistic interpretability → SAE features and their applications
The Context Access Divide: Interaction-Level Architecture as a Complementary Dimension of Agentic InequalityLLM agents and tool use → General LLM agent frameworks and benchmarks
VEGAS: Human-Aligned Video Caption Evaluation via GazeMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Conversational Retrieval and On-the-Fly Knowledge Modeling of Historical Penitentiary Repression RecordsRetrieval-augmented generation → RAG pipeline design and evaluation
Harness VLA: Steering Frozen VLAs into Reliable Manipulation Primitives via Memory-Guided AgentsLLM agents and tool use → Multimodal and embodied agents
Statistical Efficiency and Inference of Quantile Distributional Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
ADORN: Adaptive Drift handling for Open RAN using Reinforcement LearningDeep reinforcement learning methods → RL for energy, networks and operations
DeltaV: Thinking with Visual State Updates in Unified Large Multimodal ModelsLLM reasoning → Multimodal reasoning +1
OmniFood-Bench: Evaluating VLMs for Nutrient Reasoning and Personalized Health AdviceLLM reasoning → Clinical and biomedical applications +1
When Synthetic Speech Is All You Have: Better Call GRPOLLM reasoning → Reinforcement learning for reasoning
Game Theory Driven Multi-Agent Framework Mitigates Language Model HallucinationLLM agents and tool use → General LLM agent frameworks and benchmarks
TRACE: A Two-Channel Robust Attribution Watermark via Complementary Embeddings for LLM-Agent TrajectoriesLLM agents and tool use → Agent threats, guardrails and governance +1
Prompt Compression via Activation AggregationEfficient LLM inference and serving → Prompt and context compression
Token-Flow Firewall: Semantic Runtime Auditing for Persistent AI AgentsLLM agents and tool use → Agent threats, guardrails and governance
Joint Discrete-Continuous Flow Matching for Open-Vocabulary Inverse Design of Multilayer Optical CoatingsDiffusion models → Flow matching, time series and science
WCog-VLA: A Dual-Level World-Cognitive Vision-Language-Action Model for End-to-End Autonomous DrivingVision-language-action models → VLA for autonomous driving +1
MobiDiff: Semantic-Aware Multi-Channel Discrete Diffusion for Human Mobility Data GenerationDiffusion models → Diffusion and flow policies for robotics
Certified Interventional Fidelity: Anytime-Valid, Adaptive Evaluation of Causal Claims in Mechanistic InterpretabilityMechanistic interpretability → Circuits and mechanistic analysis
AnyDexRT: Calibration-Free Dexterous Hand Retargeting with Few-Shot Human GuidanceOffline RL and imitation learning from data → Robot manipulation imitation learning
Spectral Analysis of Dueling Q-LearningDeep reinforcement learning methods → Deep RL algorithms and representations
AutoAnchor: Stable Diffusion Unlearning Using Cross-Attention as a Manifold SurrogateDiffusion models → Robustness, privacy and adversarial aspects +1
GitLake: Git-for-data for the agentic lakehouseLLM agents and tool use → Software engineering coding agents
Write-Protected Discrete Bottlenecks for Language-Grounded World Models: A Structural Limitation and Sufficient FixWorld models → World models for language agents
From Legacy Documentation to OSCAL: An MCP-Based Agent Pipeline for Threat-Informed Continuous Compliance in Critical InfrastructureLLM agents and tool use → Agent threats, guardrails and governance
Psychological Competence as a Missing Dimension in AI EvaluationLLM reasoning → Reasoning benchmarks and evaluation
Multi-Agent Firewall Architecture for Privacy Protection of Sensitive Data in Interactions with Language ModelsLLM agents and tool use → Agent threats, guardrails and governance
Deep Reinforcement Learning-Empowered Wireless Sensor Networking for 6G Closed-Loop ControlsDeep reinforcement learning methods → RL for energy, networks and operations
PolyUQuest: Verifiable Structure-Aware Web RAG over Heterogeneous GraphsRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
On the Design of Mixture-of-Experts for Dynamic Gaussian SplattingMixture-of-experts → Vision mixture-of-experts
Closing the Null Space: Guidance-Aware Quantization for Classifier-Free DiffusionDiffusion models → Diffusion foundations and sampling
Playing ZendoWorld: Challenging AI Agents on Active Visual Concept InductionLLM agents and tool use → Multimodal and embodied agents
On the Limitations of Non-GPU AI Accelerators for Large-Model Inference: A Field Study of MoE and Multimodal Serving on Huawei AscendMixture-of-experts → Efficient MoE training and inference systems
A First-Principles Theory of Slow Thinking and Active PerceptionLLM reasoning → Test-time compute and efficient thinking
Open-ended Multi-agent Autocurricula via Visual Inspection of Policies with Multi-modal LLMsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
LEEVLA: Seeing What Matters in Latent Environment Evolution for Vision-Language-ActionVision-language-action models → World model-based VLA
Out of Sight: Compression-Aware Content Protection against Agentic CrawlersLLM reasoning → Knowledge- and tool-augmented reasoning +2
ASMR: Agentic Schema Generation for Ship Maintenance Report WritingLLM agents and tool use → Web, search, and research agents
Overthinking: Amplifying Reasoning Weights to Extract Learned SecretsLLM reasoning → Test-time compute and efficient thinking
SQuaD-SQL: Efficient Text-to-SQL with Small Language Models via LLM-Guided Knowledge DistillationKnowledge distillation → Knowledge distillation for LLMs
Prismata: Confining Cross-Site Prompt Injection in Web AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
Securing Autonomous Vehicle Systems via Twin-Aware Federated Reinforcement LearningEnd-to-end autonomous driving → Adversarial scenarios and safety testing +1
Answer Set Programming Energised! End-to-End Neurosymbolic Reasoning and Learning with ASP and Energy Based ModelsLLM reasoning → Reasoning benchmarks and evaluation
Understanding and Mitigating the Video-Action Generalization Gap via Temporal RatioWorld models → Embodied and robot world models
TTHE: Test-Time Harness EvolutionLLM agents and tool use → Software engineering coding agents
CausalDS: Benchmarking Causal Reasoning in Data-Science AgentsLLM reasoning → Causal reasoning in LLMs +1
Mixture of Enhanced-View Experts for Multi-Query Vehicle ReID and A Large-Scale BenchmarkMixture-of-experts → Vision mixture-of-experts
MASTE: A Multi-Agent Pipeline for Zero-Shot Aspect Sentiment Triplet ExtractionLLM agents and tool use → General LLM agent frameworks and benchmarks
Post-Training in End-to-End Autonomous DrivingEnd-to-end autonomous driving → End-to-end system architecture
Persuasion Attacks Can Decrease Effectiveness of CoT MonitoringLLM reasoning → Prompting and chain-of-thought reasoning
When LLMs Agree, Are They Right? Auditing Self-Consistency and Cross-Model Agreement as Confidence SignalsLLM reasoning → Test-time compute and efficient thinking +1
When Thinking Hurts: Epistemic Signals in the Reasoning Chains of Visual Language ModelsLLM reasoning → Multimodal reasoning +1
Towards Efficient Large Language Model Serving: A Survey on System-Aware KV Cache OptimizationEfficient LLM inference and serving → KV cache optimization
Reinforcing the Generation Order of Multimodal Masked Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
What LLM Forecasters Know but Don't Say: Probing Internal Representations for Calibration and FaithfulnessLLM reasoning → Time series reasoning
RadioDiff-v2: Generative Angular Radio Maps for Multi-Beam Selection and LocalizationDiffusion models → Diffusion for compression and wireless
An exact information theory of generalization phase transitions in Bayesian diffusion modelsDiffusion models → Diffusion foundations and sampling
PLURAL: A Global Dataset for Value AlignmentLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment

All 344