Papers

New on arXiv

Fri 21 Aug 2026

365 papers

Schr"odinger's Code Repository: Have LLMs Learned SWE-bench or Memorized It?LLM agents and tool use → Software engineering coding agents
RelCheck: Dual-Evidence Spatial Grounding for VLM Hallucination CorrectionMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
MechReason: Benchmarking Multi-Image Multi-Hop Reasoning in Mechanical EngineeringLLM reasoning → Multimodal reasoning +1
Test-Time Scaling for Scientific Equation DiscoveryLLM reasoning → Test-time compute and efficient thinking
RedLight-VLA: Models for traffic-rule grounding and behavioral emphasis in driving policiesEnd-to-end autonomous driving → Reasoning VLA and VLM driving policies
REFINE: A Multi-Agent LLM Approach for Evidence-Guided Code RefactoringLLM agents and tool use → Software engineering coding agents
From Traceability to Justifiability: Accountability Structures in Agentic Software EngineeringLLM agents and tool use → Agent threats, guardrails and governance
Context as an Environment: Programmatic Context Management for Long-Horizon AgentsLLM agents and tool use → Agent memory systems
SketchFlow: Zero-Shot Vector Sketch Generation via GMM Prior Flow in CLIP Latent SpaceDiffusion models → Image generation, text-to-image and editing
Mitigating Database Leakage in RAG Systems with Keyword-Grounded Fact SubstitutionRetrieval-augmented generation → Privacy and security of RAG
FlowSem: Flow Matching for Adaptive Wireless Image Transmission in Semantic CommunicationDiffusion models → Diffusion for compression and wireless
ExploreAI: Agentic Exploration Knowledge Bases for Reproducible Observable-Regression Testing of Black-Box VR and 3D ApplicationsLLM agents and tool use → Multimodal and embodied agents
SAEM: Stage-Aware Expert Management for Memory-Efficient MoE Inference in Chain-of-Thought ReasoningLLM reasoning → Test-time compute and efficient thinking +1
Semantic Compression Trees: Multi-Resolution Knowledge Retrieval via Hierarchical Semantic ResidualsRetrieval-augmented generation → Generation and context integration
Can LLMs Truly Forget? Revealing Unlearning Gaps Through Adversarial EvaluationLLM safety and security → Machine unlearning
Exploring Agentic Approaches for Data Issue Detection and Repair in AI-Assisted VisualizationLLM agents and tool use → Multimodal and embodied agents
K-Bench: measuring model performance on real scientific agent requestsLLM agents and tool use → Web, search, and research agents
Perturb the Thought, Not the Pixels: Latent-Space Rollout Diversification for Reinforcement Learning of Vision-Language ModelsLLM reasoning → Multimodal reasoning +1
Reading the Room: Implicit Confusion Encoding in Recurrent World Model StatesWorld models → Latent predictive world models
Anchoring Bias: A Persistent Fairness Backdoor Attack against MLLMs under Continual LearningMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Automating Multi-Hop RAG Evaluation via TRIAD: From Context Extraction to Validated Dataset GenerationRetrieval-augmented generation → RAG pipeline design and evaluation
Forgotten in Weights, Recovered by Tools: Agentic Tool Unlearning for LLM AgentsLLM agents and tool use → Tool learning and function calling +1
Beyond Sparse Weights: When Is Attention Compressible?Efficient LLM inference and serving → KV cache optimization
DamageScope: Vision-Language Retrieval at Scale for Disaster Damage Assessment from Satellite ImageryMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Neuro-Formal Verification: Agentic Language-Agnostic Formal Program ReasoningLLM reasoning → Reasoning benchmarks and evaluation +1
SecOPD: Mitigating Adaptive Prompt Injections by On-Policy DistillationLLM safety and security → Prompt injection and agent security
TASSO: TAsk-Specific Subspace Optimization for Continual Learning of Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Structural Inference in Undocumented Mobile Databases: A Reproducible Benchmark for Evaluating Agentic Reasoning in Digital ForensicsLLM agents and tool use → Web, search, and research agents
Asymmetric Capacity Allocation in Self-Refinement PipelinesLLM agents and tool use → General LLM agent frameworks and benchmarks
Natural-Language Workflows Are Not Software Yet: Artifact-Driven Compilation for Reliable Agent ExecutionLLM agents and tool use → Software engineering coding agents
AI-to-AI Code Reviews of GitHub Pull RequestsLLM agents and tool use → Software engineering coding agents
Beyond Fault Localization: A Trajectory-Level Study of LLM Agents for Microservice Root Cause AnalysisLLM agents and tool use → Software engineering coding agents
Rethinking Expressivity and Efficiency in Test-Time TrainingLLM reasoning → Test-time compute and efficient thinking
Re3Cap: Retrieval-Guided Refinement for Image Captioning Enhancement via Reinforcement LearningLLM reasoning → Multimodal reasoning +1
Difficulty-Calibrated Interpolation Paths for Conditional Flow MatchingDiffusion models → Flow matching, time series and science
CLEAR: Continuous Latent Adapter Routing for Utility-Preserving LLM Safety AlignmentLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Memory Augmentation Unlocks Efficient Chain-of-Thought ReasoningLLM reasoning → Test-time compute and efficient thinking
EnSI-RAG: Entity-Structure-Indexed Retrieval-Augmented Generation for Long-Document Question AnsweringRetrieval-augmented generation → RAG robustness and adaptive retrieval
Just Noticeable Difference Modeling for Token Compression in Vision-Language-Action ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +2
SPICE: Speculative Prefetching with Low-Rank Expert Surrogates and Heterogeneous Orchestration for MoE Inference AccelerationMixture-of-experts → Efficient MoE training and inference systems +1
RARE: Decoupling Representation Steering from Expert Routing in Mixture-of-Experts Language ModelsMixture-of-experts → Expert routing strategies
Beyond Attention Masks: Instruction Anchoring for Efficient In-Context Diffusion GenerationDiffusion models → Diffusion foundations and sampling
Personalized Privacy Control in LLMs via Attention Head InterventionLLM agents and tool use → Agent threats, guardrails and governance
SlimDiffuSE: Towards Efficient Diffusion-Based Speech Enhancement using Slimmable NetworksDiffusion models → Audio, speech and music
Is Visual Prompting All You Need? Studying VLM Spatial Reasoning under Progressive Visual ScaffoldsLLM reasoning → Spatial and embodied reasoning +1
Graph Engineering in the Era of LLM Agents: From Individual Intelligence to System IntelligenceLLM agents and tool use → General LLM agent frameworks and benchmarks
Llama-Mobile: Efficient 2.7-Bit Quantization of VLMsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
TraceGrant: A Contract-Governed Security Framework for the Task-Effect Lifecycle of Networked LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
CIVA: Critic-Induced Value-Subspace Attacks on Visual World-Model AgentsWorld models → World model security and GUI agents
ClawSentry: A Progressive Multi-Tier Security Monitor for Safeguarding Autonomous LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
ReFrame: Evidence-Guided Test-Time Safety Alignment in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +2
Trustworthy RAG: An Evaluation Agent for Detecting Misinformation and Knowledge Poisoning in Generative AI SystemsRetrieval-augmented generation → Privacy and security of RAG +1
Designing a Robust LLM-Based Evaluation System for Agentic AI in Drug Discovery Through Human AlignmentLLM agents and tool use → Medical and scientific domain agents
Z2-ACT: End-to-End Verifiable Agentic Intent Control for Open 6G RANLLM agents and tool use → Agentic AI for wireless networks
Roadside-Cooperative Autonomous Driving: From Data Platform to Vision-Language End-to-End ReasoningLLM reasoning → Chain-of-thought reasoning for driving +1
COMET: Contrastive Motion-Enhanced Temporal Reasoning for Video Multimodal Large Language ModelsLLM reasoning → Multimodal reasoning +1
Don't Solve, Just Compare: Tiny Advisors for Runtime Intervention in LLM AgentsLLM agents and tool use → Agent skill learning and evolution
Recognition-Conditioned Reasoning: A Training-Free Multimodal-LLM Pipeline for Fine-Grained Micro-Action UnderstandingLLM reasoning → Spatial and embodied reasoning
Latent Ordinal Evidence, Misaligned Outputs: Inference-Time Ordinal Lens Alignment for Multimodal LLMsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Trojaning the Alignment: Stealthy Backdoor Attacks against Graph Foundation ModelsLLM safety and security → Backdoor and data poisoning attacks
Belief Without Behavior: Measuring the Translation of Theory of Mind into Coordinated Social Action in Vision-Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
WA-JEPA: Rethinking the Video JEPA Paradigm for World-Action Modeling in Autonomous DrivingWorld models → World models for autonomous driving +1
TreeWY: Speculative Verification for Gated DeltaNet HybridsEfficient LLM inference and serving → Speculative decoding
Can Scientific Claims Be Removed from Large Language Models? A Systematic Evaluation of Claim-Level UnlearningLLM safety and security → Machine unlearning
Neural-Primitive: An Efficient End-to-end Local Planner with Primitive-based Imitation Learning for Autonomous FlightOffline RL and imitation learning from data → Imitation learning methods
The Logic of Machine Self-PreservationLLM agents and tool use → General LLM agent frameworks and benchmarks
No Judgment Without a Reason: Counterfactual Receipts for Versioned AI EvaluatorsLLM agents and tool use → Agent threats, guardrails and governance
Graph-Operator World Models for Morphology-Parameter Generalization in Continuous ControlWorld models → Embodied and robot world models
Decoupling Policy Extraction for Offline Reinforcement LearningOffline RL and imitation learning from data → Offline RL algorithms
Generation of Web Apps with Agentic IDEs: An Empirical AssessmentLLM agents and tool use → Software engineering coding agents
A Collaborative Multi-Modality Interaction for VLA-based End-to-End Autonomous DrivingVision-language-action models → VLA for autonomous driving +1
KREL: Automatic Medical Coding via Knowledge-Guided Reasoning over Clinical Evidence with LLMsLLM reasoning → Clinical and biomedical applications
ReCurveflow: A Flow Matching Framework that Learns Curved Reaction Trajectories to Predict Transition State GeometriesDiffusion models → Flow matching, time series and science
BC-Bench: Evaluating Agentic Engineering in a Domain-Specific Language for ERPLLM agents and tool use → General LLM agent frameworks and benchmarks
RAG Deserves an Index: Why Ingest-Time Compilation Beats Query-Time InterpretationRetrieval-augmented generation → RAG robustness and adaptive retrieval
TRACE: Agentic Catalog Enrichment with Multi-source Evidence GroundingLLM agents and tool use → Social simulation and multi-agent debate
SAC-Copula: Quality-Preserving Watermarking for Diffusion Language Models via Smooth Correlated Gumbel FieldsDiffusion models → Robustness, privacy and adversarial aspects +1
Certified Multi-Turn Robustness for LLM Safety via Compositional Bounds and Safety PersistenceLLM safety and security → Jailbreaking
Scaling Muon for Diffusion TransformersDiffusion models → Diffusion foundations and sampling
Enhancing Localized Reasoning for Long Video Understanding via Efficient Segment-to-Video SupervisionLLM reasoning → Multimodal reasoning +1
Automated Trajectory Evaluation for Mobile Agents via Step-Level Consequence Reasoning and AggregationLLM agents and tool use → Agent threats, guardrails and governance
CertVLA: Certified Defense against Physical Visual Attacks for Vision-Language-Action ModelsVision-language-action models → Safety and adversarial attacks
Structure for Reading, Prose for Writing: Asymmetric Structural Conditioning in Multi-Agent Document AuthoringLLM agents and tool use → Web, search, and research agents
Rethinking Demonstration Unlearning in Imitation Learning for RoboticsOffline RL and imitation learning from data → Imitation learning methods +1
Tree-of-Concerns: Hierarchical Multi-Agent Debate for Unstated-Limitation Extraction in Scientific CritiqueLLM agents and tool use → Social simulation and multi-agent debate
CAS: Conformalized Agentic Search via Adaptive Retrieval and Policy WeightingLLM agents and tool use → Web, search, and research agents
CARD: Diagnosing Belief to Action Routing Failures in Vision Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Fuzzy-MoE: Interpretable Regime-Conditioned Expert Routing for Non-Stationary Multivariate Time Series ForecastingMixture-of-experts → MoE for time series and graphs
Vis-Poison: Poisoning Visual Knowledge in Multimodal Retrieval-Augmented GenerationMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Identity-Preserving Text-to-Video Generation via Agentic Enhancement and Semantic RepairLLM agents and tool use → Multimodal and embodied agents
Is Multimodal Speculative Decoding Ready for Diffusion-Based Parallel Drafting? A Survey and Empirical DiagnosisEfficient LLM inference and serving → Fast autoregressive visual generation
ForeTime-VLA: Causal Future-Token Distillation from a World Action Model for Conveyor-Belt ManipulationWorld models → Embodied and robot world models +1
Calibrating Criterion Revision in LLM Agents: Failure Modes and a Trace-Anchored ProtocolLLM agents and tool use → General LLM agent frameworks and benchmarks
AffordAny: Open-World 3D Affordance Grounding from Monocular RGB Images via Vision-Language-Guided Geometric ReasoningLLM reasoning → Spatial and embodied reasoning
DirEAG: Dirichlet Evidence Aggregation for Calibrating Verbalized Confidence in Mathematical ReasoningLLM reasoning → Test-time compute and efficient thinking
ArtiMo: Agent-Driven Articulated Mesh AnimationLLM agents and tool use → Multimodal and embodied agents
Temporal Validity on Real Software Histories: Eliminating Stale-Fact Errors in Code-Assistant Memory over GitHub FixesRetrieval-augmented generation → RAG for code
Why2Speak: Faithful Reasoning for Abstaining Action PoliciesLLM agents and tool use → General LLM agent frameworks and benchmarks
Auditable by Construction: An Ontology-Driven Framework for Trustworthy LLM Analytics in Enterprise FinanceRetrieval-augmented generation → Legal and financial RAG
The Claws in Plain Sight: Unauthorized Context Disclosure through LLM Agent Tool CallsLLM agents and tool use → Agent threats, guardrails and governance

All 365