Papers

New on arXiv

Sun 24 May 2026

263 papers

SonicSampler: Unified Tile-Aware Kernels for LLM Sampling and Speculative VerificationEfficient LLM inference and serving → Speculative decoding
Incomplete Prompt Jailbreaks in Large Language ModelsLLM safety and security → Jailbreaking
Robust Critics: Defending LLMs Against Multi-Turn AttacksLLM safety and security → Safety alignment, refusal and guardrails
Catalyst Papers in Artificial Intelligence Research: A Landscape on ICLR from 2017 to 2025LLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Specifying AI-SDLC Processes: A Protocol Language for Human-Agent BoundariesLLM agents and tool use → Agent threats, guardrails and governance
DLLM-JEPA: Joint Embedding Predictive Architectures for Masked Diffusion Language ModelsDiffusion models → Discrete and text diffusion
Fine-Tuning Vision-Language Models for Understanding Current Damage and Scoring Priority with Quality Guard AgentMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Tool-Schema Compression Enables Agentic RAG Under Constrained Context BudgetsRetrieval-augmented generation → RAG pipeline design and evaluation +1
Furina: Fragmented Uncertainty-Driven Refusal Instability AttackLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
When Does Adaptive Guidance Help? Belief-Aware Privileged Distillation for Autonomous Driving Under Partial ObservabilityDeep reinforcement learning methods → RL for autonomous driving and traffic
MemMorph: Tool Hijacking in LLM Agents via Memory PoisoningLLM agents and tool use → Agent threats, guardrails and governance
Geometry-Aware Image Flow MatchingDiffusion models → Flow matching, time series and science
READER: Reasoning-Enhanced AI-Generated Text DetectionLLM reasoning → Reasoning benchmarks and evaluation
Latent Q-Barrier Shielding for Safe In-Context Reinforcement LearningDeep reinforcement learning methods → Safe and constrained RL
Inference Time Optimization with Confidence DynamicsLLM reasoning → Test-time compute and efficient thinking
Transformer Field Theory: A Response-Theoretic Approach to Mechanistic InterpretabilityMechanistic interpretability → Circuits and mechanistic analysis
Multi-Objective Learning for Diffusion Models: A Statistical Theory under Semi-Supervised LearningDiffusion models → Diffusion foundations and sampling
GroupTravelBench: Benchmarking LLM Agents on Multi-Person Travel PlanningLLM agents and tool use → General LLM agent frameworks and benchmarks
Hide to Guide: Learning via Semantic MaskingLLM reasoning → Reinforcement learning for reasoning
Baton: Explicit Semantic Blueprints for Joint Video-Audio GenerationDiffusion models → Audio, speech and music
Localize and Neutralize: Gradient-guided Token Suppression against Visual Prompt Injection AttackLLM safety and security → Vision-language and text-to-image safety
Injecting Image Guidance into Text-Conditioned Diffusion Models at InferenceDiffusion models → Image generation, text-to-image and editing
Directional Alignment Mitigates Reward Hacking in Reinforcement Learning for Language ModelsLLM alignment / RLHF / preference optimisation → Reward hacking and misspecification
DarkForest: Less Talk, Higher Accuracy for Multi-Agent LLMsLLM reasoning → Reasoning for agents, planning and code +1
By Their Fruits You Will Know Them: Comparing Formalizations of Law by the Decisions They EncodeLLM reasoning → Reasoning for agents, planning and code
Knowledge Graph-Driven Expert-Level Reasoning for NeuroscienceLLM reasoning → Knowledge- and tool-augmented reasoning
Locality Matters for Training-Free Audio Token Compression in Audio-Language ModelsEfficient LLM inference and serving → Audio and omni-modal token compression
Grow-Prune-Freeze Networks: Adaptive & Continual Learning Technique for Olfactory NavigationWorld models → Latent dynamics models and model-based RL
AME-TS: Anchored Mixture-of-Experts for Time Series ForecastingMixture-of-experts → MoE for time series and graphs
ScaleWoB: Guiding GUI Agents with Coding Agents via Large-Scale Environmental SynthesisLLM agents and tool use → GUI and mobile agents
Beyond the Frontier: Stochastic Backtracking for Efficient Test-Time ScalingLLM reasoning → Test-time compute and efficient thinking
LLM Agent Based Renewable Energy Forecasting Using Edge and IoT Data A Review of Solar Wind Weather and Grid Aware Decision SupportLLM agents and tool use → Financial trading and forecasting agents
Blocked Gibbs meets Diffusion Transformers: Unsupervised Learning for Constraint OptimizationDiffusion models → Diffusion foundations and sampling
Inference-Time Alignment of Diffusion Models via Trust-Region Iterative Twisted Sequential Monte CarloDiffusion models → Diffusion foundations and sampling
Counterfactually Safe Reinforcement LearningDeep reinforcement learning methods → Safe and constrained RL
Multi-Agent Specification-based Metamorphic Testing of FMU-Based SimulationsLLM agents and tool use → Software engineering coding agents
Polynomial Context-Truncation Sensitivity in Autoregressive Language Models: Sequential Wyner-Ziv Bounds for KV Cache CompressionEfficient LLM inference and serving → KV cache optimization
WorldCraft: From Camera Navigation to Object Manipulation in Interactive Video World ModelsWorld models → Video-generation world models and simulators
Security in the Fine-Tuning Lifecycle of Large Language Models: Threats, Defenses,Evaluation, and Future DirectionsLLM safety and security → Safety alignment, refusal and guardrails
Reinforcement Learning with a Bilevel World-Model Architecture for Scan-Order Optimisation in Laser Directed Energy DepositionWorld models → Latent dynamics models and model-based RL +1
Cultivating Machine Intelligence: The OMEGA Shift from Top-Down Optimization to Autopoietic Cognitive EcologiesLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
X-DiffVLA: X-Embodied Diffusion Action Heads for Vision-Language-Action ModelsVision-language-action models → Human-video and latent-action pretraining
AstroRAG -- A Pagerank-Based Retrieval-Augmented Generation Pipeline for Question Answering in AstronomyRetrieval-augmented generation → Agentic RAG, agents and memory
Language Bias in LVLMs: From In-Depth Analysis to Simple and Effective MitigationMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning +1
MimirRAG: A Multi-Agent RAG Framework for Financial Data Retrieval with Metadata IntegrationRetrieval-augmented generation → Legal and financial RAG +1
ParkingWorld: End-to-End Autonomous Parking Reinforcement Learning from Corrective Experience in 3DGS SimulationDeep reinforcement learning methods → RL for autonomous driving and traffic
D3S2: Diffusion-Guided Dataset Distillation for Semantic SegmentationDiffusion models → Medical imaging and segmentation
Stop Denoising Your BlursDiffusion models → Inverse problems and image restoration
A perspective on fluid mechanical environments for challenges in reinforcement learningDeep reinforcement learning methods → RL theory and sample complexity
Meta-Modal Agent: Sequential Evidence Routing for Missing-Modality Candidate RerankingLLM agents and tool use → Agentic recommender systems
MemMark: State-Evolution Attribution Watermarking for Agent Long-Term Memory SystemsLLM agents and tool use → Agent threats, guardrails and governance +1
Better, Faster: Harnessing Self-Improvement in Large Reasoning ModelsLLM reasoning → Test-time compute and efficient thinking
NeurIPS: Neuro-anatomical Inductive Priors for Sphere-based Brain DecodingMixture-of-experts → Medical and biomedical MoE
Selective Test-Time Compute Scaling for Click-Through Rate Prediction via Uncertainty-Triggered Feature Path ExplorationLLM reasoning → Test-time compute and efficient thinking
Self-Balancing Gradient Allocation for Heterogeneity-Aware Feature Generation in Click-Through Rate PredictionDiffusion models → Diffusion foundations and sampling
Universal Boosts, Specific Suppressors: Sparse Autoencoder Steering of Medical Vision-Language ModelsMultimodal LLMs / vision-language models → Medical vision-language models +1
Bridging the Gap: Enabling Soft Actor Critic for High Performance Legged LocomotionDeep reinforcement learning methods → RL for robot locomotion and manipulation
Investigating the Interplay between Contextual and Parametric Chain-of-Thought Faithfulness under OptimizationLLM reasoning → Prompting and chain-of-thought reasoning
Mitigating Object Hallucinations in Vision-Language Models through Region-Aware Attention RecalibrationMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
NITP: Next Implicit Token Prediction for LLM Pre-trainingMixture-of-experts → MoE large language models
APT-Agent: Automated Penetration Testing using Large Language ModelsLLM agents and tool use → Agent threats, guardrails and governance
Interpretability Transfer from Language to Vision via Sparse AutoencodersMechanistic interpretability → Interpreting vision and multimodal models
Memory-Induced Tool-Drift in LLM AgentsLLM agents and tool use → Agent memory systems
Global linear convergence of entropy-regularized softmax policy gradient beyond tabular MDPsDeep reinforcement learning methods → RL theory and sample complexity
MambaDSF: Multi-Scale SSM with Dilated Feature Fusion for Sonar Small Target DetectionState-space sequence models (S4, Mamba) → Point clouds, detection and remote sensing
Dynamic Neural Koopman Distillation for Real-Time Robot Control Using Diffusion ModelsDiffusion models → Diffusion and flow policies for robotics
Factorize to Generalize: Retrieval-Guided Invariant-Dynamic Decomposition for Time Series ForecastingRetrieval-augmented generation → Time series and agricultural RAG
When Reasoning Hurts: Source-Aware Evaluation of Frontier LLMs for Clinical SOAP Note GenerationLLM reasoning → Clinical and biomedical applications
ProActor: Timing-Aware Reinforcement Learning for Proactive Task Scheduling AgentsDeep reinforcement learning methods → RL training of language and agent models
X-Foresight: A Joint Vision-Action Causal Forecasting Network via Predictive World ModelingWorld models → Embodied and robot world models +1
QuoVLA: Quotient Space for Vision-Language-Action ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Towards a Universal Causal ReasonerLLM reasoning → Causal reasoning in LLMs
Trajectory-Consistent Calibration for Cache-Accelerated Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Clustering as Reasoning: A k-Means Interpretation of Chain-of-Thought Graph LearningLLM reasoning → Prompting and chain-of-thought reasoning
Unifying Value Alignment and Assignment in Cross-Domain Offline Reinforcement Learning with Heterogeneous DatasetsOffline RL and imitation learning from data → Offline RL algorithms
Geo-Expert: Towards Expert-Level Geological Reasoning via Parameter-Efficient Fine-TuningLLM reasoning → Spatial and embodied reasoning
Reflect-Guard: Enhancing LLM Safeguards against Adversarial Prompts via Logical Self-ReflectionLLM safety and security → Safety alignment, refusal and guardrails
Optimus: Elastic Decoding for Efficient Diffusion LLM ServingEfficient LLM inference and serving → Serving systems and request scheduling
Test-Time Deep Thinking to Explore Implicit RulesLLM reasoning → Test-time compute and efficient thinking
Agent Manufacturing: Foundation-Model Agents as First-Class Industrial EntitiesLLM agents and tool use → General LLM agent frameworks and benchmarks
RouteScan: A Non-Intrusive Approach to Auditing MoE LLMs Safety via Expert Routing TelemetryMixture-of-experts → Expert routing strategies
Cross-Domain Energy-Guided Diffusion Generation for Off-Dynamics Reinforcement LearningDiffusion models → Diffusion and flow policies for robotics +2
Divide-and-Conquer Inference for Large-Scale Visual Recognition with Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
DUEL: Adversarial Self-Play for Multimodal ReasoningLLM reasoning → Multimodal reasoning +2
Beyond the Target: From Imitation to Collaboration in Speculative DecodingEfficient LLM inference and serving → Speculative decoding
CONF-KV: Confidence-Aware KV Cache Eviction with Mixed-Precision Storage for Long-Horizon LLMEfficient LLM inference and serving → KV cache optimization
PANDO: Efficient Multimodal AI Agents via Online Skill DistillationLLM agents and tool use → Web, search, and research agents
Seeing Without Understanding: Large Language Model Evaluation of Mobile User Interface Quality, Failure Taxonomy, and Architectural Explanation
Exploring the Role of LLMs in HPC Programming: A Survey
Safe Evolution with Circuit Anchors
VeriSimpl: Robust Optimization Modeling from Natural Language using Simplification-based Verification
OverFlowLight: Real-Time Gridlock Prevention and Traffic Signal Optimization for Urban Intersections
Analysis and Prediction of At-Risk Students Using Machine Learning Algorithms
A Geometric Gaussian Mixture Representation of Plane Curves
Samudra 2: Scaling Ocean Emulators across Resolutions
Aligning Cellular Sheaves with Classifier Attention for Interpretable Weakly-Supervised Pathology Localization
Gait2Hip-60: A Unified Deep Learning Benchmark for Predicting Hip Muscle Forces and Joint Moments from Multi-Cadence Gait Kinematics
From Affect to Complex Behavior: Advancing Multimodal Human-Centered AI at the 10th ABAW Workshop & Competition
Context Features Are Cheap: Rank-Aware Decomposition for Efficient Feature Interaction in Recommender Systems
Checking Fact with Better Retrieval: Dynamic Contrastive Learning for Evidence Retrieval

All 263