Papers

New on arXiv

Thu 23 Apr 2026

360 papers

The Coin Flip Judge? Reliability and Bias in LLM-as-a-Judge EvaluationLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
ProcessThinker: Enhancing Multi-modal Large Language Models Reasoning via Rollout-based Process RewardLLM reasoning → Multimodal reasoning +2
How Consistent Are LLM Agents? Measuring Behavioral Reproducibility in Multi-Step Tool-Calling PipelinesLLM agents and tool use → Tool learning and function calling
SketchVLM: Vision language models can annotate images to explain thoughts and guide usersLLM reasoning → Multimodal reasoning +1
When Policies Cannot Be Retrained: A Unified Closed-Form View of Post-Training Steering in Offline Reinforcement LearningOffline RL and imitation learning from data → Offline RL algorithms
AutoRISE: Agent-Driven Strategy Evolution for Red-Teaming Large Language ModelsLLM agents and tool use → Agent threats, guardrails and governance +1
Probing Visual Planning in Image Editing ModelsLLM reasoning → Multimodal reasoning
IntrAgent: An LLM Agent for Content-Grounded Information Retrieval through Literature ReviewLLM agents and tool use → Web, search, and research agents
Emergent Strategic Reasoning Risks in AI: A Taxonomy-Driven Evaluation FrameworkLLM reasoning → Reasoning for agents, planning and code
PermaFrost-Attack: Stealth Pretraining Seeding(SPS) for planting Logic Landmines During LLM TrainingLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
An End-to-End Ukrainian RAG for Local Deployment. Optimized Hybrid Search and Lightweight GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
Memanto: Typed Semantic Memory with Information-Theoretic Retrieval for Long-Horizon AgentsLLM agents and tool use → Agent memory systems
Insect-inspired modular architectures as inductive biases for reinforcement learningDeep reinforcement learning methods → RL for robot navigation and UAV control
Sound Agentic Science Requires Adversarial ExperimentsLLM agents and tool use → Web, search, and research agents
PrivUn: Unveiling Latent Ripple Effects and Shallow Forgetting in Privacy UnlearningLLM safety and security → Machine unlearning
Outcome Rewards Do Not Guarantee Verifiable or Causally Important ReasoningLLM reasoning → Reinforcement learning for reasoning
Incentivizing Neuro-symbolic Language-based Reasoning in VLMs via Reinforcement LearningLLM reasoning → Reasoning benchmarks and evaluation
Lightweight Retrieval-Augmented Generation and Large Language Model-Based Modeling for Scalable Patient-Trial MatchingRetrieval-augmented generation → Medical and biomedical RAG
Robust Localization for Autonomous Vehicles in Highway ScenesEnd-to-end autonomous driving → End-to-end system architecture
Source-Modality Monitoring in Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Read the Paper, Write the Code: Agentic Reproduction of Social-Science ResultsLLM agents and tool use → Web, search, and research agents
Conditional Diffusion Posterior Alignment for Sparse-View CT ReconstructionDiffusion models → Medical imaging and segmentation
MambaCSP: Hybrid-Attention State Space Models for Hardware-Efficient Channel State PredictionState-space sequence models (S4, Mamba) → Core SSM architectures for language
Long-Horizon Manipulation via Trace-Conditioned VLA PlanningVision-language-action models → Reasoning and chain-of-thought in VLA
MathDuels: A Self-Play Benchmark That GrowsSynthetic data and self-improvement for LLMs → Self-play RL for reasoning
When Prompts Override Vision: Prompt-Induced Hallucinations in LVLMsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs +1
From Research Question to Scientific Workflow: Leveraging Agentic AI for Science AutomationLLM agents and tool use → Web, search, and research agents
A Scale-Adaptive Framework for Joint Spatiotemporal Super-Resolution with Diffusion ModelsDiffusion models → Inverse problems and image restoration
Nemobot Games: Crafting Strategic AI Gaming Agents for Interactive Learning with Large Language ModelsLLM agents and tool use → Reinforcement learning for LLM agents
Replay-buffer engineering for noise-aware quantum circuit optimizationDeep reinforcement learning methods → Quantum reinforcement learning
Transient Turn Injection: Exposing Stateless Multi-Turn Vulnerabilities in Large Language ModelsLLM safety and security → Jailbreaking
Causality-Encoded Diffusion Models for Interventional Sampling and Edge InferenceDiffusion models → Diffusion foundations and sampling
Black-Box Skill Stealing Attack from Proprietary LLM Agents: An Empirical StudyLLM agents and tool use → Agent threats, guardrails and governance
Tool Attention Is All You Need: Dynamic Tool Gating and Lazy Schema Loading for Eliminating the MCP/Tools Tax in Scalable Agentic WorkflowsLLM agents and tool use → Tool learning and function calling
Quotient-Space Diffusion ModelsDiffusion models → Flow matching, time series and science
Learning to Communicate: Toward End-to-End Optimization of Multi-Agent Language SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
Agentic AI-Enabled Framework for Thermal Comfort and Building Energy Assessment in Tropical Urban NeighborhoodsLLM agents and tool use → Financial trading and forecasting agents
Thinking with Reasoning Skills: Fewer Tokens, More AccuracyLLM reasoning → Test-time compute and efficient thinking
StructMem: Structured Memory for Long-Horizon Behavior in LLMsLLM reasoning → Reasoning for agents, planning and code
Agentic AI-assisted coding offers a unique opportunity to instill epistemic grounding during software developmentLLM agents and tool use → Web, search, and research agents
Hi-WM: Human-in-the-World-Model for Scalable Robot Post-TrainingWorld models → Embodied and robot world models +1
Ramen: Robust Test-Time Adaptation of Vision-Language Models with Active Sample SelectionMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
AEL: Evolving Agent Harness in Open-Ended EnvironmentsLLM agents and tool use → Agent memory systems
Stealthy Backdoor Attacks against LLMs Based on Natural Style TriggersLLM safety and security → Backdoor and data poisoning attacks
WorldMark: A Unified Benchmark Suite for Interactive Video World ModelsWorld models → Video-generation world models and simulators
Encoder-Free Human Motion Understanding via Structured Motion DescriptionsLLM reasoning → Spatial and embodied reasoning
Task-specific Subnetwork Discovery in Reinforcement Learning for Autonomous Underwater NavigationDeep reinforcement learning methods → RL for robot navigation and UAV control
To See the Unseen: on the Generalization Ability of Transformers in Symbolic ReasoningLLM reasoning → Reasoning benchmarks and evaluation
DCMorph: Face Morphing via Dual-Stream Cross-Attention DiffusionDiffusion models → Robustness, privacy and adversarial aspects
Process Supervision via Verbal Critique Improves Reasoning in Large Language ModelsLLM reasoning → Test-time compute and efficient thinking
Language as a Latent Variable for Reasoning OptimizationLLM reasoning → Reasoning benchmarks and evaluation
Sculpt4D: Generating 4D Shapes via Sparse-Attention Diffusion TransformersDiffusion models → 3D, point cloud and scene generation
AgenticQwen: Training Small Agentic Language Models with Dual Data Flywheels for Industrial-Scale Tool UseLLM agents and tool use → Reinforcement learning for LLM agents
CoFEE: Reasoning Control for LLM-Based Feature DiscoveryLLM reasoning → Reasoning benchmarks and evaluation
Separable Expert Architecture: Toward Privacy-Preserving LLM Personalization via Composable Adapters and Deletable User ProxiesLLM safety and security → Machine unlearning
Pre-trained LLMs Meet Sequential Recommenders: Efficient User-Centric Knowledge DistillationKnowledge distillation → Distillation for recommender systems
Seeing Isn't Believing: Uncovering Blind Spots in Evaluator Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
DiffNR: Diffusion-Enhanced Neural Representation Optimization for Sparse-View 3D Tomographic ReconstructionDiffusion models → Medical imaging and segmentation
OptiVerse: A Comprehensive Benchmark towards Optimization Problem SolvingLLM reasoning → Test-time compute and efficient thinking
GeoMind: An Agentic Workflow for Lithology Classification with Reasoned Tool InvocationLLM agents and tool use → Multimodal and embodied agents
Efficient Agent Evaluation via Diversity-Guided User SimulationLLM agents and tool use → General LLM agent frameworks and benchmarks
Frozen LLMs as Map-Aware Spatio-Temporal Reasoners for Vehicle Trajectory PredictionLLM reasoning → Chain-of-thought reasoning for driving
Rethinking Cross-Domain Evaluation for Face Forgery Detection with Semantic Fine-grained Alignment and Mixture-of-ExpertsMixture-of-experts → Vision mixture-of-experts
MCP Pitfall Lab: Exposing Developer Pitfalls in MCP Tool Server Security under Multi-Vector AttacksLLM agents and tool use → Agent threats, guardrails and governance
Dynamical Priors as a Training Objective in Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Do MLLMs Understand Pointing? Benchmarking and Enhancing Referential Reasoning in Egocentric VisionLLM reasoning → Spatial and embodied reasoning
The Privacy Guardian Agent: Towards Trustworthy AI Privacy AgentsLLM agents and tool use → Agent threats, guardrails and governance
Reasoning Primitives in Hybrid and Non-Hybrid LLMs: Do Architectural Differences Yield Advantages in State-Tracking and Recall?LLM reasoning → Test-time compute and efficient thinking
AI-Gram: When Visual Agents Interact in a Social NetworkLLM agents and tool use → Social simulation and multi-agent debate
FairQE: Multi-Agent Framework for Mitigating Gender Bias in Translation Quality EstimationLLM agents and tool use → Social simulation and multi-agent debate
CSC: Turning the Adversary's Poison against ItselfLLM safety and security → Backdoor and data poisoning attacks
S1-VL: Scientific Multimodal Reasoning Model with Thinking-with-ImagesLLM reasoning → Multimodal reasoning
A Task Decomposition and Planning Framework for Efficient LLM Inference in AI-Enabled WiFi-Offload NetworksEfficient LLM inference and serving → Serving systems and request scheduling
VG-CoT: Towards Trustworthy Visual Reasoning via Grounded Chain-of-ThoughtLLM reasoning → Multimodal reasoning +1
From Noise to Intent: Anchoring Generative VLA Policies with Residual BridgesVision-language-action models → Generalist VLA models and benchmarks
Conjecture and Inquiry: Quantifying Software Performance Requirements via Interactive Retrieval-Augmented Preference ElicitationRetrieval-augmented generation → RAG pipeline design and evaluation
VLAA-GUI: Knowing When to Stop, Recover, and Search, A Modular Framework for GUI AutomationLLM agents and tool use → GUI and mobile agents
Prototype-Based Test-Time Adaptation of Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
ReaGeo: Reasoning-Enhanced End-to-End Geocoding with LLMsLLM reasoning → Spatial and embodied reasoning
Learn Weightlessness: Imitate Non-Self-Stabilizing Motions on Humanoid RobotDeep reinforcement learning methods → RL for robot locomotion and manipulation
Symbolic Grounding Reveals Representational Bottlenecks in Abstract Visual ReasoningLLM reasoning → Multimodal reasoning
Latent Denoising Improves Visual Alignment in Large Multimodal ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Sub-Token Routing for KV Cache CompressionEfficient LLM inference and serving → KV cache optimization
Ideological Bias in LLMs' Economic Causal ReasoningLLM reasoning → Causal reasoning in LLMs
Teacher-Guided Routing for Sparse Vision Mixture-of-ExpertsMixture-of-experts → Vision mixture-of-experts
Understanding and Mitigating Spurious Signal Amplification in Test-Time Reinforcement Learning for Math ReasoningLLM reasoning → Reinforcement learning for reasoning
CI-Work: Benchmarking Contextual Integrity in Enterprise LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
PaperMind: Benchmarking Agentic Reasoning and Critique over Scientific Papers in Multimodal LLMsLLM reasoning → Knowledge- and tool-augmented reasoning +1
Strategic Heterogeneous Multi-Agent Architecture for Cost-Effective Code Vulnerability DetectionLLM agents and tool use → Agent threats, guardrails and governance
LatRef-Diff: Latent and Reference-Guided Diffusion for Facial Attribute Editing and Style ManipulationDiffusion models → Image generation, text-to-image and editing
Can MLLMs "Read" What is Missing?Multimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI GroundingDeep reinforcement learning methods → RL training of language and agent models
Enhancing Online Recruitment with Category-Aware MoE and LLM-based Data AugmentationMixture-of-experts → MoE large language models
When Agents Look the Same: Quantifying Distillation-Induced Similarity in Tool-Use BehaviorsLLM agents and tool use → General LLM agent frameworks and benchmarks
Planning Beyond Text: Graph-based Reasoning for Complex Narrative GenerationLLM reasoning → Reasoning benchmarks and evaluation
CAP: Controllable Alignment Prompting for Unlearning in LLMsLLM safety and security → Machine unlearning
Reasoning About Traversability: Language-Guided Off-Road 3D Trajectory PlanningLLM reasoning → Chain-of-thought reasoning for driving +2
ReCAPA: Hierarchical Predictive Correction to Mitigate Cascading FailuresVision-language-action models → Failure recovery and robustness
SparKV: Overhead-Aware KV Cache Loading for Efficient On-Device LLM InferenceEfficient LLM inference and serving → KV cache optimization
Zero-Shot Detection of LLM-Generated Text via Implicit Reward ModelLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment

All 360