Papers

New on arXiv

Sat 8 Aug 2026

242 papers

Toward Self-Repairing Ubiquitous Robots Using Goal-Oriented Agentic AI in Human-Robot InteractionsLLM agents and tool use → General LLM agent frameworks and benchmarks
When Verifiers Vote Backwards under Verdict Substitution: Signed Pivotal Value in Correlated Self-ConsistencyLLM reasoning → Test-time compute and efficient thinking
ChainUQ: Reasoning Consistency-Aware Uncertainty Quantification for Large Language ModelsLLM reasoning → Test-time compute and efficient thinking
CricRAG: Retrieval Augmented Vision-Language Models for Personalized Cricket CoachingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
OSFoundry: Building and Evolving Operating Systems with Specification-Guided AgentsLLM agents and tool use → Software engineering coding agents
MME-Safety: A Fine-grained Benchmark for Safety Evaluation of MLLMsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Do MLLMs Really Understand Low-Resource Khmer Documents? A Pilot Study on Khmer Document VQAMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
OpRAG: A Resource-Deterministic Runtime for GPU-Backed Multi-Stage RAG WorkflowsRetrieval-augmented generation → RAG pipeline design and evaluation
Circuit Fine-Tuning for Compute-Efficient Transformer AdaptationMechanistic interpretability → Circuits and mechanistic analysis
StructReward: Efficient Structured Process Rewards for Self-Correcting Multimodal ReasoningLLM reasoning → Multimodal reasoning +1
Your VLM Already Knows When: Training-Free Temporal Grounding by Asking Yes or NoMultimodal LLMs / vision-language models → Video understanding with MLLMs
Ouroboros: A Self-Developing Frontier Coding Agent with Reviewed Core EvolutionLLM agents and tool use → Software engineering coding agents
Query-Only Backdoor Attacks on Self-Evolving Skills via Trajectory PoisoningLLM agents and tool use → Agent threats, guardrails and governance +1
A Controlled Study of Feature-Based Knowledge Distillation Across Student DesignsKnowledge distillation → General knowledge distillation methods
Stateful CARS: Exact Cross-History Reuse for Policy-Constrained LLM AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
OBLIVION: Workflow-Level Operational Skill Unlearning for Deployed AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
AraSSM: A bidirectional state-space encoder for Arabic masked language modelingState-space sequence models (S4, Mamba) → Core SSM architectures for language
Learning from Environmental Feedback: Credit Assignment across Multiple Timescales for Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents
The Replay Gap: Static Evaluation of Model Switching in LLM Agents Scores the Wrong WorldLLM agents and tool use → Software engineering coding agents
SAGE: SLO-Aware Adaptive Retrieval for Production RAG SystemsRetrieval-augmented generation → RAG robustness and adaptive retrieval
LatticeMind: A Conflict-Aware Memory Primitive for Multi-Agent SystemsLLM agents and tool use → Agent memory systems
Control-Diverse Reinforcement Fine-Tuning: Decoupling the Shared Control Bottleneck of RL Post-TrainingLLM reasoning → Reinforcement learning for reasoning
VTO: Visual Tool Orchestration for Video Anomaly DetectionDeep reinforcement learning methods → RL training of language and agent models
Harmful Content Is Not Enough: Continuation Framing Moderates In-Context Emergent MisalignmentLLM safety and security → Safety alignment, refusal and guardrails
Conditional Diffusion for Nonparametric Instrumental Variable Quantile RegressionDiffusion models → Diffusion foundations and sampling
A Minimal kappa--τ Logic for Risk-Sensitive AbductionLLM reasoning → Reasoning benchmarks and evaluation
Matching Supervision to the Student's Learning Capacity: A Unified Framework for On-Policy Self-DistillationLLM reasoning → Prompting and chain-of-thought reasoning +1
Thinking vs. NoThinking: Towards Interpreting Reasoning Mechanisms of Large Language Models via Sparse AutoencodersLLM reasoning → Reasoning benchmarks and evaluation +1
Wiener Representation Filtering for VLM Hallucination SuppressionMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
STEMMA: An Adversarial Multi-Agent Framework for Evaluating Self-Identity Consistency in LLMsLLM agents and tool use → Social simulation and multi-agent debate +1
Agentic AI-driven Immersive Simulation: A Knowledge-Aware Virtual Training Platform forHigh Dose Rate (HDR) BrachytherapyLLM agents and tool use → Medical and scientific domain agents
Can LLM Agents Stick to the Script? A Benchmark for Long-Horizon Consistency in Interactive NarrativesLLM agents and tool use → Multimodal and embodied agents
A Unified Framework for Dynamic Reward Shaping in Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Long SKILL Compliance as Logical Reasoning: Closure-Grounded Detection with Scaling-Guided On-Policy DistillationLLM reasoning → Reasoning for agents, planning and code +1
DS@GT ARC at Touch'e: Large Language Models for Retrieval-Augmented DebateRetrieval-augmented generation → RAG robustness and adaptive retrieval
Compositional Cross-Modality Translation via Whole-Volume Multitask Latent Flow MatchingDiffusion models → Medical imaging and segmentation
Compositional Threat Analysis of Latent Compromise in LLM Agent Systems: The Order 66 ScenarioLLM agents and tool use → Agent threats, guardrails and governance
Understanding Security and Privacy Perceptions of Content Creators Regarding AI Labels of AI-Generated ContentLLM safety and security → Watermarking generated content and models
Improving Constraint Models with LLM AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Think Deep, Speak Once: Relit, A Recursive Latent Implicit Transformer FrameworkLLM reasoning → Test-time compute and efficient thinking
Defending Retrieval-Augmented Intrusion Detection Against Knowledge Poisoning and Prompt InjectionRetrieval-augmented generation → Privacy and security of RAG +1
OasisKV: Scaling In-Decode KV Cache Beyond HBM with Lookahead Sparse PrefetchingEfficient LLM inference and serving → KV cache optimization
Archer: Adaptive Reuse of Cached Hidden States for Efficient Rollback in Diffusion Language ModelsDiffusion models → Discrete and text diffusion
Commitment Before Realization: When Classifier-Free Guidance Becomes Unnecessary in Masked Diffusion Language ModelsDiffusion models → Discrete and text diffusion
RotaryQuant: Fitting 120B MoE Models on Consumer Hardware via Fused Compressed-Space AttentionMixture-of-experts → MoE large language models
Explore, Map, Remember, Decide: Are Embodied VLMs Ready for Safety-Critical Scenarios?Multimodal LLMs / vision-language models → Embodied and 3D vision-language models
SurgWMBench: A Vision-Based Benchmark for World-Modeling Surgical Instrument Motion PlanningWorld models → Medical and surgical world models
NeuroGuard: Neural Gradient Update Aware of Representation DamageKnowledge distillation → Distillation for continual learning
CORDA: A Benchmark for Hierarchical Harm-Centric Moral Reasoning in Large Language ModelsLLM reasoning → Reasoning benchmarks and evaluation
ZOMP: Zeroth-Order Multi-Modal Prompt Tuning for Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
SodaMem: Evidence-Grounded Temporal Graph Memory for LLM AgentsLLM agents and tool use → Agent memory systems
PhysX-CoT: Structured Physical Reasoning from a Single Image to Simulation-Ready 3D AssetsLLM reasoning → Spatial and embodied reasoning
Lingjing: A Simulation Testbed for Multi-Agent Embodied Tasks in Open-Ended CitiesLLM agents and tool use → Multimodal and embodied agents
Stateful Multi-Agent LLMs for Cross-View Interface Alignment in Automotive Model-Based Systems EngineeringLLM agents and tool use → Software engineering coding agents
SkillSmith: Enhancing Locally Deployed Agents via Automatic Skill Construction and EvolutionLLM agents and tool use → Agent skill learning and evolution
Compiling and Benchmarking Task-State Horizons for Embodied AgentsLLM agents and tool use → Multimodal and embodied agents
Decided Upstream, Written Late: Locating and Pricing the Cross-Lingual Refusal Circuit of a Multilingual MoELLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +2
Do All LLMs Know When They're Being Harmful? A Reproducibility Study of Latent-Space Safety Probes Across Model FamiliesLLM safety and security → Safety alignment, refusal and guardrails
BASIS: Breach-Aware Selective Prompt Injection Shielding with Prefill Attention ProbesLLM safety and security → Prompt injection and agent security
Prompt Embedding Probes (PEP): Hallucination Detection in LLMs from Hidden StatesMechanistic interpretability → Hallucination detection from model internals
4D-WAM: Infusing Spatiotemporal Awareness into World Action Models through Trajectory FieldsWorld models → Embodied and robot world models
Evidence-RL: Towards Evidence-intensive Visual ReasoningLLM reasoning → Multimodal reasoning +1
Thought-Level Beam Search for ReasoningLLM reasoning → Test-time compute and efficient thinking
Ground-Truth Neighborhood Regularization for Reinforcement Learning Post-Training of Time Series Foundation ModelsDeep reinforcement learning methods → Deep RL algorithms and representations
Evidence-Grounded Forensic Reasoning for Detecting and Grounding Multi-Modal Media ManipulationLLM reasoning → Multimodal reasoning
Quality-Diversity Stress Tests for Process Reward Models:What Archive Coverage Can and Cannot CertifyLLM reasoning → Reinforcement learning for reasoning +1
Evaluator Ensembles Under Reward Hacking: Covariance Geometry and Finite-Search GuaranteesLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
PE-Mamba: Bidirectional Selective Layer Aggregation for AI-Generated Image DetectionState-space sequence models (S4, Mamba) → Point clouds, detection and remote sensing
Advantage-Guided Gate: Reshaping Open-Ended Reasoning for Vision-Based Spatial IntelligenceLLM reasoning → Spatial and embodied reasoning +1
Distilling Physical Priors into Streaming World ModelsWorld models → Latent predictive world models
Verication-driven closed-loop multi-agent large language modelframework for code-compliant structural designLLM agents and tool use → Multimodal and embodied agents
Thinking Hard, Not Smart: Reasoning Models Fail to Ration Test-Time Compute Across QuestionsLLM reasoning → Test-time compute and efficient thinking
CyberAGENTS: Structured Autonomy for Agentic Gamified Learning in CybersecurityLLM agents and tool use → Agent threats, guardrails and governance
EasyBalance: Cross-Layer Load Balancing in Distributed MoE InferenceMixture-of-experts → Efficient MoE training and inference systems
Self-Evolving Neuro-Symbolic Skills for Tool-Augmented Spatial ReasoningLLM reasoning → Spatial and embodied reasoning +2
KGCache: Amortized Subgraph Retrieval for KG Reasoning with LLMsLLM reasoning → Knowledge- and tool-augmented reasoning
Persistent Semantic Entities in Tool-Augmented LLM SystemsLLM agents and tool use → Agent memory systems
VIOLET: High-Fidelity Violin Synthesis with Techniques and DynamicsDiffusion models → Audio, speech and music
Adaptive Supervised Anchoring for On-Policy Self-DistillationKnowledge distillation → On-policy and self-distillation for LLMs
EvoTrustRAG: Evolution-Aware Conflict Attribution and Evidence Handling for Reliable Retrieval-Augmented GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
REIN: Bridging the Gap between Reasoning and Reliability via Reflection and Abstention AlignmentLLM reasoning → Test-time compute and efficient thinking
ZhuLong: Execution-Grounded LLM Agent for EDA Scripting with Offline API Self-ExplorationLLM agents and tool use → Software engineering coding agents
Forged Peer Judgments Mislead Multimodal LLM Judge Panels: Source-Blind Anchoring and Panel-Consensus VerificationMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
SPECTRA: Pushing the KV Cache Beyond the 2-Bit Cliff via Spectral Transform CodingLLM agents and tool use → Software engineering coding agents +1
Private Anytime Selective-Risk Certification for Federated Retrieval-Augmented Generation: Guarantees and Empirical LimitsRetrieval-augmented generation → Privacy and security of RAG
Reproducible Evaluation of MoE Expert Caching: Replay Semantics, Workload Contamination, and Operating RegimesMixture-of-experts → Efficient MoE training and inference systems
GraphThink: Graph-Enhanced LLM Thinking for Long-Horizon Embodied Task PlanningLLM reasoning → Spatial and embodied reasoning
TelemetrySuffBench: Is Agent Telemetry Sufficient for Failure-Origin Diagnosis?LLM agents and tool use → Software engineering coding agents
Auditing Instruction-Trajectory Mismatches in Multimodal Robot DemonstrationsVision-language-action models → Failure recovery and robustness
Capability-Routed Guard: Defending Large Reasoning Models Against Reasoning-Centric JailbreaksLLM safety and security → Safety alignment, refusal and guardrails
Router Sensitivity Under Lightweight Fine-Tuning Identifies Prunable Experts in Mixture-of-Experts ModelsMixture-of-experts → Expert routing strategies
Reason Wide, Not Deep: Amortizing the Reasoning Premium into Distilled SkillsLLM reasoning → Test-time compute and efficient thinking +1
V-Simba: Unleashing the Architectural Potential of RL in Visual Continuous ControlDeep reinforcement learning methods → RL for robot locomotion and manipulation
LHSDet: High-Resolution AI-Generated Image Detection via Visual Question AnsweringDiffusion models → Image generation, text-to-image and editing
How Much Does It Cost to Answer My Question? Benchmarking Cloud VLM-based VQA SystemsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
CommitKV: Lifecycle-Aware KV Cache Compression via Commit Transitions for Multi-Turn AgentsEfficient LLM inference and serving → Efficient reasoning and agent inference
"Many Are My Names": The Anatomy of the Assistant and Its Personas via Sparse AutoencodersMechanistic interpretability → Activation steering methods
Counterfactual Benchmarking and Training for Factuality Consistency and Order-Robust Grounded Reasoning in LLMs over Heterogeneous KnowledgeLLM reasoning → Knowledge- and tool-augmented reasoning
From Risk Scoring to Risk Allocation: A Density-Driven Framework for Diverse Monitoring in Multi-Agent Systems
TailSpec-EASE: Knowledge-Graph-Regularized Linear Recommendation for Web Long-Tail Discovery

All 242