Papers

New on arXiv

Sun 9 Aug 2026

259 papers

MGRL-RSCC: Multi-Granularity Reward Reinforcement Learning for Fine-Grained Remote Sensing Change CaptioningDeep reinforcement learning methods → RL training of language and agent models
Spectral Tail Interventions in Decoder-Only Language Models: Reasoning-Sensitive Weight Structure from Controlled SurgeryLLM reasoning → Test-time compute and efficient thinking
"As a Language Model...": Chat Template Switches LLM Self-Referential Voice and Activation Steering Reproduces ItMechanistic interpretability → Probing deception, bias and beliefs
Look Before You Leap: Pre-Action Verification for LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
SeqFeed: Improving Agentic RTL Code Generation with Sequential Behavior FeedbackLLM agents and tool use → Software engineering coding agents
Reading is not Reasoning: Bridging the Agentic Policy Gap in Vision-Text CompressionLLM agents and tool use → Agent memory systems +1
Idea Search: Guiding Tree Search with Ideas to Explore Diverse Scientific MethodsLLM reasoning → Test-time compute and efficient thinking
Topology-Aware Global-Local Mamba Networks for Palm Vein BiometricsState-space sequence models (S4, Mamba) → Mamba for medical image analysis
Independent Patch Verification for Coding Agents with a Bidirectional Reconstruct-and-Verify FrameworkLLM agents and tool use → Software engineering coding agents
What Would Fix This RAG Failure? Auditing Counterfactual Response with Paired Evidence InterventionsRetrieval-augmented generation → RAG robustness and adaptive retrieval
Difficulty-Gated Fusion of Reasoning Views for Temporal RetrievalLLM reasoning → Knowledge- and tool-augmented reasoning
Not an A11y: How Android Accessibility Exposes Mobile AI Agents to Indirect Prompt InjectionLLM safety and security → Prompt injection and agent security
Integrated Multimodal AI System for Retrieval-Augmented Reasoning, Object Sensing, and Damage AnalysisLLM reasoning → Multimodal reasoning +1
Toward CT-Equivalent Image Quality in Low-Dose Radiotherapy Planning: Conditional Diffusion-Based CBCT-to-CT Synthesis and the Impact of CBCT Input RepresentationDiffusion models → Medical imaging and segmentation
Tied Trit-Planes: Constraining PTQTP to a Uniform Nine-Level Quantizer, with a Persistent Folded Format for Disk-Streamed Mixture-of-Experts ServingMixture-of-experts → MoE large language models
From Recovery to Drop-off: How Action Post-training Reduces a VLM's Late-Layer Depth DecodabilityMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
LLM Reasoning for Subjective Tasks: Failure Modes, Mitigation, and Dynamic Reasoning RoutingLLM reasoning → LLM reasoning for recommendation
AquiLLM: An Architecture for Supporting Tacit Knowledge Capture in Research GroupsRetrieval-augmented generation → RAG pipeline design and evaluation
Theory-Guided Deception Detection: A RAG-Based Artificial Intelligence ExplorationRetrieval-augmented generation → Hallucination detection in RAG
DistillCache: KL-Guided Adaptive KV-Cache Eviction for Memory-Efficient LLM InferenceEfficient LLM inference and serving → KV cache optimization
Beyond Routing: Decoupling Expert Dispatch and Aggregation in Sparse Mixture-of-ExpertsMixture-of-experts → Expert routing strategies
Findings of the First Teaching Monster Challenge: A Benchmark of Pedagogical Content Knowledge in AI AgentsLLM agents and tool use → Educational agents and math problem solving
SG-WAM: Text-Grounded and Spatial-aware Semantic Guidance for World-Action ModelsWorld models → Embodied and robot world models
Deployable Per-Instance Multi-Layer Activation Steering for Large Language ModelsMechanistic interpretability → Activation steering methods
Distilling Vision-Language Models for Robust Traffic Sign Perception in Autonomous VehiclesMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Tevatron-Elastic: A Unified Abstraction for Training Elastic Retrievers and RerankersEfficient LLM inference and serving → Prompt and context compression
AdapterMoE: A Two-Stage Hard-Routing Mixture-of-Experts Architecture for Multi-Crop Disease Recognition with Calibrated Rejection and Incremental LearningMixture-of-experts → Vision mixture-of-experts
Improving Generalization Robustness of Multimodal RLVRMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
IDRAAK: From Multi-Agent NLP to Few-Shot Prompting for Semantic Drift Detection in Technical RequirementsLLM agents and tool use → Software engineering coding agents
Toward Metacognitive One-Shot Indirect Prompt Injection: Strategy Abstraction Via Outcome-Conditioned ReflectionLLM safety and security → Prompt injection and agent security
Deferred Audio Pruning with Local Audio-Visual Dynamics for Omni-LLMsEfficient LLM inference and serving → Audio and omni-modal token compression
Evidence-Calibrated Runtime Reconstruction for Agent Skills Across Heterogeneous Coding AgentsLLM agents and tool use → Agent skill learning and evolution
Unsure but Certain: Uncovering the Representation-Confidence Gap in Diffusion Language ModelsDiffusion models → Discrete and text diffusion
SymDiag: Explainable Diagnosis for LLM Reasoning via Neuro-Symbolic VerificationLLM reasoning → Reasoning benchmarks and evaluation
OmnilingualGAIA2: Evaluating the Multilingual Gap in Frontier AI AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
A Mean-Field Framework for Inference-Time Distributional Control of Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Learning from Consensus and Disagreement: Unsupervised On-Policy Self-Distillation with Minority-Trajectory ContrastKnowledge distillation → On-policy and self-distillation for LLMs
A Distribution Mapping Approach to Counterfactually Fair Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
FitAQA: A Benchmark of Fitness Action Quality Assessment for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
PAST: Privileged Adaptation from Complete Student Trajectories for On-Policy Self-DistillationLLM reasoning → Prompting and chain-of-thought reasoning +1
WA-SpecDec: World-Aware Speculative Decoding for Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation
LibraSpec: Dynamic Diffusion-Based Speculative Decoding via Marginal-Gain-Driven OptimizationEfficient LLM inference and serving → Speculative decoding
High-Quality Exposure Correction with Diffusion-Based Image Generation PriorsDiffusion models → Diffusion foundations and sampling
Resolution Meets Reduction: Efficient Visual Context for 3D Radiology Report GenerationMultimodal LLMs / vision-language models → Medical vision-language models
PluginEval: A Diagnostic Benchmark for Fine-Grained Error Attribution in Function CallingLLM agents and tool use → Tool learning and function calling
A Structural Dynamics Graph World Model: Unified Modeling, Constrained Rollout, and Interpretable CalibrationWorld models → World models for language agents
RippleKV: Cross-Layer KV Cache Allocation via Perturbation PropagationEfficient LLM inference and serving → KV cache optimization
Efficient Test-Time Scaling for LLM-based Time Series ForecastingLLM reasoning → Time series reasoning
Catastrophic Forgetting in Continual Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
The Scaffolding Matters More Than the Interface: A Controlled Comparison of MCP and CLI Tool Use Across Seven Agent Scaffoldings, Five Language Models, and One Software TaskLLM agents and tool use → Software engineering coding agents
The Evolution of Mixture-of-Experts Architectures in Large Language Models: Routing, Topology, Load Balancing, and Expert ParallelismMixture-of-experts → Expert routing strategies
Agentic Visual Reasoning in Whole-Slide Pathology Images via Active PerceptionLLM reasoning → Clinical and biomedical applications
SkillReason: Reasoning-Enhanced Agent Skill Retrieval for Implicit User RequestsLLM reasoning → Reasoning for agents, planning and code +1
CuteTTS: Efficient and High-Quality Speech Synthesis via Autoregressive Modeling of Continuous LatentsDiffusion models → Audio, speech and music
VLZip: Unified Visual and Textual Compression for Interleaved Long-Context ModelingMultimodal LLMs / vision-language models → Visual token pruning and efficiency
UniMoMo: Expert Merging-Based MoE Acceleration for Large Recommendation ModelsMixture-of-experts → MoE for recommender systems
MedCalc-R1: Knowledge-Guided Reward Framework for Medical Mathematical ReasoningLLM reasoning → Clinical and biomedical applications
VADER: Adaptive Debiasing for Hallucination Mitigation in Video Large Language ModelsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Business Arena: Benchmarking LLM Agents in a Realistic MarketplaceLLM agents and tool use → Financial trading and forecasting agents
REVEAL: A Rubric-Guided Agent for Explicit Evidence Sufficiency Verificationin Long-Video Question AnsweringRetrieval-augmented generation → Multimodal RAG
ForestBench: A Unified Graph Framework for Evaluating Multi-Agent CollaborationLLM agents and tool use → General LLM agent frameworks and benchmarks
Unaccountable Delegation, Fading Skills: Mapping the Risks of Workplace AI AgentsLLM agents and tool use → Agent threats, guardrails and governance
Population-Scalable Multi-Agent World ModelingWorld models → Video-generation world models and simulators
Knowledge-Distilled End-to-End Reinforcement Learning for Smooth 6-DOF Thrust Control and Rapid Adaptation to Ocean Currents in Remotely Operated VehiclesDeep reinforcement learning methods → RL for robot navigation and UAV control
SDDBMs: Soft Denoising Diffusion Bridge ModelsDiffusion models → Diffusion foundations and sampling
VoxZip: Semantic-Anchored Temporal KV Cache Compression for Long-Context Audio InferenceEfficient LLM inference and serving → Audio and omni-modal token compression
Vid2WAM: Distilling Video Diffusion Priors into World Action ModelsWorld models → Embodied and robot world models +1
SC-Diff: Semantically Calibrated Diffusion for Visible-to-Infrared Image TranslationDiffusion models → Image generation, text-to-image and editing
MotionCraft: Latent World Modeling with Sparse Attention for Visual UpscalingWorld models → Video-generation world models and simulators
When Skills Meet Safety: Benchmarking and Characterizing the Adaptive Jailbreak Robustness of Skill-Merged LLMsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Can Graph Learning Learn Circuits?Mechanistic interpretability → Circuits and mechanistic analysis
eBIRD: Event-based Intensity Image Reconstruction Using Controllable Diffusion ModelsDiffusion models → Video generation
Fluid Structure, Rigid Record: A Layered Organizational Design Framework for Agent-Native OrganizationsLLM agents and tool use → General LLM agent frameworks and benchmarks
Reproducing and Stress-Testing Two Approaches to LLM Reasoning Reliability: Test-Time Probability Aggregation and Logic-Representation EditingLLM reasoning → Reasoning benchmarks and evaluation
MathShikkha: A Controlled Study of Answer-Only and Chain-of-Thought Supervision for Bangla Mathematical Reasoning in Small Language ModelsLLM reasoning → Mathematical reasoning
SocialFiVis: A Visual Analytics Sandbox for LLM-Grounded Multi-Agent Simulation in Social FinanceLLM agents and tool use → Social simulation and multi-agent debate
Linguistically-Aligned and Visually-Grounded Preference Optimization for Clinically-Augmented Medical Report GenerationLLM alignment / RLHF / preference optimisation → Medical LLM alignment
TrustRoboReward: Preference-Ordered Isotonic Score Editing for Multi-Paradigm Robot Reward ModelsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
HoloAegis: Frozen Representation, Topological Inference --- Minimally Parametric Safety Manifolds and Their Capability Boundaries for LLM GuardrailsLLM safety and security → Safety alignment, refusal and guardrails
FlashBoot: Sub-Second Weight Loading for Large Models at Rack ScaleMixture-of-experts → Efficient MoE training and inference systems
VectraYX-Vision-1B: A Sub-2B Spanish/LATAM Cybersecurity Vision-Language Model with Structured Visual Reasoning and Native Tool UseLLM reasoning → Reasoning for agents, planning and code +1
Yesterday's Shield, Today's Spear: A Self-Evolving Safety Guardrail in ProductionSynthetic data and self-improvement for LLMs → Safety of self-evolving systems +1
Aero Realtime: Fully Aligned Input-Output Streams for Low-Latency Streaming Multimodal GenerationEfficient LLM inference and serving → Audio and omni-modal token compression
SkillsMetric: Mapping the Detection Boundary of Static Analysis for Malicious Agent SkillsLLM agents and tool use → Agent threats, guardrails and governance
LLM within MCP Matters: Measuring Inefficient Resource Utilization Driven by LLMsLLM agents and tool use → Tool learning and function calling
Hierarchical Self-Improvement: A Framework for Task-Specific Evolvable Agent HarnessesLLM agents and tool use → General LLM agent frameworks and benchmarks +1
What Keeps Agent Skills from Being Reusable? Evidence from 138K SKILL.md FilesLLM agents and tool use → Agent skill learning and evolution
Calling the Bluff: Detecting Ever-Shifting Harmful Chat Dialogue via Ordered Reasoning Chain RegularizationLLM reasoning → Reasoning benchmarks and evaluation
Hidden Language Consistency Phenomena in Reasoning LLMsLLM reasoning → Reasoning benchmarks and evaluation
Forgotten History or Test-of-Time? Retrospect and Prospect on RAG from an IR PerspectiveRetrieval-augmented generation → RAG pipeline design and evaluation
Learning Deep Modality-Shared Self-Expressiveness for Image Clustering with Textual InformationMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Personalized Communication Skills for Agentic Recommender SystemsLLM agents and tool use → Agentic recommender systems
Tangent: An Empirical Study of Testing Practices for LLM-Based Agent ApplicationsLLM agents and tool use → Software engineering coding agents
From Product Search to Preference Articulation: The Economics of Agentic CommerceLLM agents and tool use → Social simulation and multi-agent debate
CAP: A Scalable Benchmark for Evaluating Cross-Site Browser Agents with Complex Actions and PerceptionLLM agents and tool use → Web, search, and research agents
Not Worth Another Token: Marginal Value Estimation for Efficient Deep Research AgentsLLM agents and tool use → Web, search, and research agents
Safety Cost of Steering Vectors Is Separable and ReducibleMechanistic interpretability → Steering and probing for LLM safety
LLMVisor: A Real-Time Latency Attribution Model for Multi-Tenant LLM ServingEfficient LLM inference and serving → Serving systems and request scheduling
Component Type, Not Reconstruction Error, Predicts Attention Quantization Sensitivity
Moving6DPoSe: A Multimodal Database for Monocular 6D Pose Estimation and Segmentation of Moving Objects

All 259