Papers

New on arXiv

Tue 28 Apr 2026

370 papers

From Agent Failure Paths to Quantified Residual Risk: A Compositional Framework for Resilient Agentic AILLM agents and tool use → Agent threats, guardrails and governance
Granularity-Regulated Adaptive Computational Efficiency for Optimal Verification in Test-Time ScalingLLM reasoning → Test-time compute and efficient thinking
Density Field State Space Models: 1-Bit Distillation, Efficient Inference, and Knowledge Organization in Mamba-2State-space sequence models (S4, Mamba) → Core SSM architectures for language
DRIVE: Modeling Skills at the Reasoning and Interaction Levels for Web Agents under Continual LearningLLM agents and tool use → Web, search, and research agents
The Safety-Aware Denoiser for Text Diffusion ModelsDiffusion models → Robustness, privacy and adversarial aspects
RA-CMF: Region-Adaptive Conditional MeanFlow for CT Image ReconstructionDeep reinforcement learning methods → RL for clinical treatment and surgery
Simple Self-Conditioning Adaptation for Masked Diffusion ModelsDiffusion models → Discrete and text diffusion
Budget-Constrained Online Retrieval-Augmented Generation: The Chunk-as-a-Service ModelRetrieval-augmented generation → RAG pipeline design and evaluation
Lifting Embodied World Models for Planning and ControlWorld models → Latent predictive world models
CacheRAG: A Semantic Caching System for Retrieval-Augmented Generation in Knowledge Graph Question AnsweringRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
Entropy Centroids as Intrinsic Rewards for Test-Time ScalingLLM reasoning → Test-time compute and efficient thinking +2
Test-Time Safety AlignmentLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Reinforcement Learning for Public Safety Power Shutoffs Under Decision-Dependent Uncertainty and Nonlinear Wildfire Ignition ModelsDeep reinforcement learning methods → RL for energy, networks and operations
Mixture of Experts Framework in Machine Learning Interatomic Potentials for Atomistic SimulationsMixture-of-experts → MoE theory and gating functions
reward-lens: A Mechanistic Interpretability Library for Reward ModelsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
Application of Deep Reinforcement Learning to Event-Triggered Control for Networked Artificial Pancreas SystemsDeep reinforcement learning methods → Spiking and neuromorphic RL
AMMA: A Multi-Chiplet Memory-Centric Architecture for Low-Latency 1M Context Attention ServingEfficient LLM inference and serving → KV cache optimization
SWE-Edit: Rethinking Code Editing for Efficient SWE-AgentLLM reasoning → Reasoning for agents, planning and code +1
Distill-Belief: Closed-Loop Inverse Source Localization and Characterization in Physical FieldsKnowledge distillation → General knowledge distillation methods
Operating-Layer Controls for Onchain Language-Model Agents Under Real CapitalLLM agents and tool use → Financial trading and forecasting agents
PPG-Based Affect Recognition with Long-Range Deep Models: A Measurement-Driven Comparison of CNN, Transformer, and Mamba ArchitecturesState-space sequence models (S4, Mamba) → Physiological signal modeling
FlowS: One-Step Motion Prediction via Local Transport ConditioningDiffusion models → Flow matching, time series and science
RaMP: Runtime-Aware Megakernel Polymorphism for Mixture-of-ExpertsMixture-of-experts → Efficient MoE training and inference systems
Training Computer Use Agents to Assess the Usability of Graphical User InterfacesLLM agents and tool use → GUI and mobile agents
Rethinking KV Cache Eviction via a Unified Information-Theoretic ObjectiveLLM reasoning → Test-time compute and efficient thinking +1
Recursive Multi-Agent SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
DV-World: Benchmarking Data Visualization Agents in Real-World ScenariosLLM agents and tool use → Multimodal and embodied agents
How Fast Should a Model Commit to Supervision? Training Reasoning Models on the Tsallis Loss ContinuumLLM reasoning → Reinforcement learning for reasoning
Pythia: Exploiting Workflow Predictability for Efficient Agent-Native LLM ServingLLM agents and tool use → Software engineering coding agents +1
TSN-Affinity: Similarity-Driven Parameter Reuse for Continual Offline Reinforcement LearningOffline RL and imitation learning from data → Decision transformers and sequence modeling
Three Models of RLHF Annotation: Extension, Evidence, and AuthorityLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
QCalEval: Benchmarking Vision-Language Models for Quantum Calibration Plot UnderstandingMultimodal LLMs / vision-language models → Visual token pruning and efficiency
When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy GradientLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
Why Are Some Emotions Harder for LLMs? Uncovering the Causal Mechanisms of Emotion Inference via Sparse AutoencodersMechanistic interpretability → Activation steering methods
Privileged Foresight Distillation: Zero-Cost Future Correction for World Action ModelsWorld models → Embodied and robot world models
SIEVES: Selective Prediction Generalizes through Visual Evidence ScoringMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent HarnessesLLM agents and tool use → Software engineering coding agents
A Distributionally Robust Reinforcement Learning Framework for Constrained Urban EV DispatchDeep reinforcement learning methods → RL for energy, networks and operations
From Soliloquy to Agora: Memory-Enhanced LLM Agents with Decentralized Debate for Optimization ModelingLLM agents and tool use → General LLM agent frameworks and benchmarks
Towards Agentic Investigation of Security AlertsLLM agents and tool use → Agent threats, guardrails and governance
Mutual Forcing: Dual-Mode Self-Evolution for Fast Autoregressive Audio-Video Character GenerationKnowledge distillation → Speech and audio distillation
Instruction-Evidence Contrastive Dual-Stream Decoding for Grounded Vision-Language ReasoningLLM reasoning → Multimodal reasoning +1
Barriers to Universal Reasoning With Transformers (And How to Overcome Them)LLM reasoning → Prompting and chain-of-thought reasoning
Improving Diversity in Black-box Few-shot Knowledge DistillationKnowledge distillation → General knowledge distillation methods
Diverse Image Priors for Black-box Data-free Knowledge DistillationKnowledge distillation → General knowledge distillation methods
Subliminal Steering: Stronger Encoding of Hidden SignalsMechanistic interpretability → Activation steering methods
Sustained Gradient Alignment Mediates Subliminal Learning in a Multi-Step Setting: Evidence from MNIST Auxiliary Logit Distillation ExperimentKnowledge distillation → On-policy and self-distillation for LLMs
SpecFed: Accelerating Federated LLM Inference with Speculative Decoding and Compressed TransmissionEfficient LLM inference and serving → Speculative decoding
SAFEdit: Does Multi-Agent Decomposition Resolve the Reliability Challenges of Instructed Code Editing?LLM agents and tool use → Software engineering coding agents
Cross-Lingual Jailbreak Detection via Semantic CodebooksLLM safety and security → Jailbreaking
Backtranslation Augmented Direct Preference Optimization for Neural Machine TranslationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
RADD: Retrieval-Augmented Discrete Diffusion for Multi-Modal Knowledge Graph CompletionRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG +1
Think Before You Act -- A Neurocognitive Governance Model for Autonomous AI AgentsLLM reasoning → Reasoning for agents, planning and code +1
K-CARE: Knowledge-driven Symmetrical Contextual Anchoring and Analogical Prototype Reasoning for E-commerce RelevanceLLM reasoning → Knowledge- and tool-augmented reasoning
CORAL: Adaptive Retrieval Loop for Culturally-Aligned Multilingual RAGRetrieval-augmented generation → Generation and context integration
Progressing beyond Art Masterpieces or Touristic Clich'es: how to assess your LLMs for cultural alignment?LLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Prefill-Time Intervention for Mitigating Hallucination in Large Vision-Language ModelsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
OxyGent: Making Multi-Agent Systems Modular, Observable, and Evolvable via Oxy AbstractionLLM agents and tool use → Software engineering coding agents
Marco-MoE: Open Multilingual Mixture-of-Expert Language Models with Efficient UpcyclingMixture-of-experts → MoE large language models
Vision SmolMamba: Spike-Guided Token Pruning for Energy-Efficient Spiking State-Space Vision ModelsEfficient LLM inference and serving → Token pruning and merging for ViTs
SnapGuard: Lightweight Prompt Injection Detection for Screenshot-Based Web AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
From CRUD to Autonomous Agents: Formal Validation and Zero-Trust Security for Semantic Gateways in AI-Native Enterprise SystemsLLM agents and tool use → Agent threats, guardrails and governance
Sample-efficient Neuro-symbolic Proximal Policy OptimizationDeep reinforcement learning methods → Interpretable and logic-guided RL
The Surprising Effectiveness of Canonical Knowledge Distillation for Semantic SegmentationKnowledge distillation → Detection and dense prediction distillation
Dyna-Style Safety Augmented Reinforcement Learning: Staying Safe in the Face of UncertaintyDeep reinforcement learning methods → Safe and constrained RL
Improving Zero-Shot Offline RL via Behavioral Task SamplingOffline RL and imitation learning from data → Offline RL algorithms
Navigating Global AI Regulation: A Multi-Jurisdictional Retrieval-Augmented Generation SystemRetrieval-augmented generation → Legal and financial RAG
One Refiner to Unlock Them All: Inference-Time Reasoning Elicitation via Reinforcement Query RefinementLLM reasoning → Reinforcement learning for reasoning
A Systematic Post-Train Framework for Video GenerationDiffusion models → Video generation
FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge DevicesEfficient LLM inference and serving → Serving systems and request scheduling
JURY-RL: Votes Propose, Proofs Dispose for Label-Free RLVRLLM reasoning → Reinforcement learning for reasoning
Biased Dreams: Limitations to Epistemic Uncertainty Quantification in Latent Dynamics ModelsDeep reinforcement learning methods → RL theory and sample complexity
CoRE: A Fine-Grained Code Reasoning Benchmark Beyond Output PredictionLLM reasoning → Reasoning for agents, planning and code
Benchmarking and Improving GUI Agents in High-Dynamic EnvironmentsLLM agents and tool use → GUI and mobile agents +1
Safe-Support Q-Learning: Learning without Unsafe ExplorationDeep reinforcement learning methods → Safe and constrained RL
PhyloSDF: Phylogenetically-Conditioned Neural Generation of 3D Skull Morphology via Residual Flow MatchingDiffusion models → Medical imaging and segmentation
Multi-action Tangled Program Graphs for Multi-task Reinforcement Learning with Continuous ControlDeep reinforcement learning methods → Deep RL algorithms and representations
Benchmarking Layout-Guided Diffusion Models through Unified Semantic-Spatial Evaluation in Closed and Open SettingsDiffusion models → Image generation, text-to-image and editing
Plausible but Wrong: A case study on Agentic Failures in Astrophysical WorkflowsLLM agents and tool use → Web, search, and research agents
ProDrive: Proactive Planning for Autonomous Driving via Ego-Environment Co-EvolutionWorld models → World models for autonomous driving +1
AHASD: Asynchronous Heterogeneous Architecture for LLM Adaptive Drafting Speculative Decoding on Mobile DevicesEfficient LLM inference and serving → Speculative decoding
Edge-Cloud Collaborative Reconstruction via Structure-Aware Latent Diffusion for Downstream Remote Sensing PerceptionDiffusion models → Inverse problems and image restoration
Cutscene Agent: An LLM Agent Framework for Automated 3D Cutscene GenerationLLM agents and tool use → Multimodal and embodied agents
Faithfulness-QA: A Counterfactual Entity Substitution Dataset for Training Context-Faithful RAG ModelsRetrieval-augmented generation → RAG robustness and adaptive retrieval
The Thinking Pixel: Recursive Sparse Reasoning in Multimodal Diffusion LatentsLLM reasoning → Multimodal reasoning +1
Exploring Time Conditioning in Diffusion Generative Models from Disjoint Noisy Data ManifoldsDiffusion models → Diffusion foundations and sampling
Combating Visual Neglect and Semantic Drift in Large Multimodal Models for Enhanced Cross-Modal RetrievalMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
MARD: A Multi-Agent Framework for Robust Android Malware DetectionLLM agents and tool use → Agent threats, guardrails and governance
DGLight: DQN-Guided GRPO Fine-Tuning of Large Language Models for Traffic Signal ControlLLM reasoning → Chain-of-thought reasoning for driving
AutoResearchBench: Benchmarking AI Agents on Complex Scientific Literature DiscoveryLLM agents and tool use → Web, search, and research agents
R-CoT: A Reasoning-Layer Watermark via Redundant Chain-of-Thought in Large Language ModelsLLM safety and security → Watermarking generated content and models
VLM Judges Can Rank but Cannot Score: Task-Dependent Uncertainty in Multimodal EvaluationMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
DRAGON: A Benchmark for Evidence-Grounded Visual Reasoning over DiagramsLLM reasoning → Multimodal reasoning
BARRED: Synthetic Training of Custom Policy Guardrails via Asymmetric DebateLLM safety and security → Safety alignment, refusal and guardrails
How Can Reinforcement Learning Achieve Expert-level Placement?Deep reinforcement learning methods → RL for energy, networks and operations
CroSearch-R1: Better Leveraging Cross-lingual Knowledge for Retrieval-Augmented GenerationRetrieval-augmented generation → Agentic RAG, agents and memory
Conditional Flow Matching for Probabilistic Downscaling of Maximum 3-day Snowfall in AlaskaDiffusion models → Flow matching, time series and science
From Insight to Action: A Novel Framework for Interpretability-Guided Data Selection in Large Language ModelsMechanistic interpretability → Knowledge and representations inside LLMs
FAMA: Failure-Aware Meta-Agentic Framework for Open-Source LLMs in Interactive Tool Use EnvironmentsLLM agents and tool use → Tool learning and function calling
ResetEdit: Precise Text-guided Editing of Generated Image via Resettable Starting LatentDiffusion models → Image generation, text-to-image and editing

All 370