Papers

New on arXiv

Wed 6 May 2026

469 papers

Semalith v1.4: A Calibrated 184M Safety Classifier Achieving State-of-the-Art Prompt-Injection Detection at 44x Fewer Parameters than Llama-Guard-3-8BLLM agents and tool use → Agent threats, guardrails and governance +1
Latent Communication Between Language Model Agents: Channels, Alignment, and the Limits of TextLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Multi-Turn Distributed Inference with Mixture of Experts for 6G Edge--Cloud NetworksMixture-of-experts → Efficient MoE training and inference systems
Role-Based Agentic AI for Intent-Driven Network and Service OrchestrationLLM agents and tool use → Agentic AI for wireless networks
SODE: Analyzing Social Dynamics in LLM AgentsLLM agents and tool use → Social simulation and multi-agent debate
NoisyCoconut: Counterfactual Consensus via Latent Space ReasoningLLM reasoning → Test-time compute and efficient thinking
Deep Dreams Are Made of This: Visualizing Monosemantic Features in Diffusion ModelsDiffusion models → Diffusion foundations and sampling +1
Test-Time Training for Visual Foresight Vision-Language-Action ModelsVision-language-action models → World model-based VLA
Beyond Penalization: Diffusion-based Out-of-Distribution Detection and Selective Regularization in Offline Reinforcement LearningOffline RL and imitation learning from data → Diffusion and flow policies for offline RL
Agentic AI and the Industrialization of Cyber Offense: Forecast, Consequences, and Defensive Priorities for Enterprises and the MittelstandLLM agents and tool use → Agent threats, guardrails and governance
Visual Text Compression as Measure TransportMultimodal LLMs / vision-language models → Visual token pruning and efficiency
Bayesian Rain Field Reconstruction using Commercial Microwave Links and Diffusion Model PriorsDiffusion models → Flow matching, time series and science
Non-Myopic Active Feature Acquisition via Pathwise Policy GradientsDeep reinforcement learning methods → RL theory and sample complexity
WAAA! Web Adversaries Against Agentic BrowsersLLM agents and tool use → Agent threats, guardrails and governance
Chainwash: Multi-Step Rewriting Attacks on Diffusion Language Model WatermarksDiffusion models → Robustness, privacy and adversarial aspects +1
FoodCHA: Multi-Modal LLM Agent for Fine-Grained Food AnalysisMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
ReaComp: Compiling LLM Reasoning into Symbolic Solvers for Efficient Program SynthesisLLM reasoning → Reasoning for agents, planning and code
Approximate Next Policy Sampling: Replacing Conservative Target Policy Updates in Deep RLDeep reinforcement learning methods → RL theory and sample complexity
LANTERN: LLM-Augmented Neurosymbolic Transfer with Experience-Gated Reasoning NetworksLLM reasoning → Reinforcement learning for reasoning
Nitsum: Serving Tiered LLM Requests with Adaptive Tensor ParallelismEfficient LLM inference and serving → Serving systems and request scheduling
Privacy Without Losing Place: A Paradigm for Private Retrieval in Spatial RAGsRetrieval-augmented generation → Privacy and security of RAG
SLAM: Structural Linguistic Activation Marking for Language ModelsLLM safety and security → Watermarking generated content and models
Authorization Propagation in Multi-Agent AI Systems: Identity Governance as InfrastructureLLM agents and tool use → Agent threats, guardrails and governance
On Semantic Loss Fine-Tuning Approach for Preventing Model Collapse in Causal ReasoningLLM reasoning → Causal reasoning in LLMs +1
The Refusal--Compliance Tradeoff: A Large-Scale Safety Behavior Audit of Large Language ModelsLLM safety and security → Safety alignment, refusal and guardrails
Information Theoretic Adversarial Training of Large Language ModelsLLM safety and security → Jailbreaking
From History to State: Constant-Context Skill Learning for LLM AgentsLLM agents and tool use → Agent skill learning and evolution
Creative Robot Tool Use by Counterfactual ReasoningLLM reasoning → Spatial and embodied reasoning
Agentic Retrieval-Augmented Generation for Financial Document Question AnsweringRetrieval-augmented generation → Legal and financial RAG +1
PRISM: Perception Reasoning Interleaved for Sequential Decision MakingLLM reasoning → Multimodal reasoning
Mise en Place for Agentic Coding: Deliberate Preparation as Context Engineering MethodologyLLM agents and tool use → Software engineering coding agents
Conditional Diffusion Under Linear Constraints: Langevin Mixing and Information-Theoretic GuaranteesDiffusion models → Inverse problems and image restoration
BALAR : A Bayesian Agentic Loop for Active ReasoningLLM reasoning → Reasoning benchmarks and evaluation +1
Partial Evidence Bench: Benchmarking Authorization-Limited Evidence in Agentic SystemsLLM agents and tool use → Agent threats, guardrails and governance
Two Steps Are All You Need: Efficient 3D Point Cloud Anomaly Detection with Consistency ModelsDiffusion models → Flow matching, time series and science
ZAYA1-8B Technical ReportLLM reasoning → Test-time compute and efficient thinking +1
Balancing Stability and Plasticity in Sequentially Trained Early-Exiting Neural NetworksEfficient LLM inference and serving → Early exit and edge inference
Counterargument for Critical Thinking as Judged by AI and HumansLLM reasoning → Reasoning benchmarks and evaluation
Feature Starvation as Geometric Instability in Sparse AutoencodersMechanistic interpretability → Sparse autoencoder training and evaluation
How Far Are VLMs from Privacy Awareness in the Physical World? An Empirical StudyMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Securing the Agent: Vendor-Neutral, Multitenant Enterprise Retrieval and Tool UseRetrieval-augmented generation → Agentic RAG, agents and memory +1
Expert Routing for Communication-Efficient MoE via Finite Expert BanksMixture-of-experts → Expert routing strategies
GLiNER Guard: Unified Encoder Family for Production LLM Safety and PrivacyLLM safety and security → Safety alignment, refusal and guardrails
Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning +1
SecureMCP: A Policy-Enforced LLM Data Access Framework for AIoT Systems via Model Context ProtocolLLM safety and security → Prompt injection and agent security
PARNESS: A Paper Harness for End-to-End Automated Scientific Research with Dynamic Workflows, Full-Text Indexing, and Cross-Run Knowledge AccumulationLLM agents and tool use → Web, search, and research agents
Career-Aware Resume Tailoring via Multi-Source Retrieval-Augmented Generation with Provenance Tracking: A Case StudyRetrieval-augmented generation → RAG robustness and adaptive retrieval
Taming Outlier Tokens in Diffusion TransformersDiffusion models → Diffusion foundations and sampling
D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion ModelsDiffusion models → Diffusion foundations and sampling
LoViF 2026 The First Challenge on Holistic Quality Assessment for 4D World Model (PhyScore)World models → Latent predictive world models
When Life Gives You BC, Make Q-functions: Extracting Q-values from Behavior Cloning for On-Robot Reinforcement LearningOffline RL and imitation learning from data → Offline RL algorithms +1
Design Conductor 2.0: An agent builds a TurboQuant inference accelerator in 80 hoursLLM agents and tool use → Software engineering coding agents
The First Token Knows: Single-Decode Confidence for Hallucination DetectionLLM reasoning → Hallucination detection and mitigation
Wasserstein-Aligned Localisation for VLM-Based Distributional OOD Detection in Medical ImagingMultimodal LLMs / vision-language models → Medical vision-language models
PSK at SemEval-2026 Task 9: Multilingual Polarization Detection Using Ensemble Gemma Models with Synthetic Data AugmentationSynthetic data and self-improvement for LLMs → LLM synthetic data generation
Superposition Is Not Necessary: A Mechanistic Interpretability Analysis of Transformer Representations for Time Series ForecastingMechanistic interpretability → Interpreting time-series and weather models
Executable World Models for ARC-AGI-3 in the Era of Coding AgentsWorld models → World models for language agents +1
ConsisVLA-4D: Advancing Spatiotemporal Consistency in Efficient 3D-Perception and 4D-Reasoning for Robotic ManipulationLLM reasoning → Spatial and embodied reasoning +1
Adaptive Policy Selection and Fine-Tuning under Interaction Budgets for Offline-to-Online Reinforcement LearningOffline RL and imitation learning from data → Offline RL algorithms
Beyond Semantics: An Evidential Reasoning-Aware Multi-View Learning Framework for Trustworthy Mental Health PredictionLLM reasoning → Clinical and biomedical applications
On the Hardness of Junking LLMsLLM safety and security → Jailbreaking
Rollout Pass-Rate Control: Steering Binary-Reward RL Toward Its Most Informative RegimeLLM agents and tool use → Reinforcement learning for LLM agents
LineRides: Line-Guided Reinforcement Learning for Bicycle Robot StuntsDeep reinforcement learning methods → RL for robot locomotion and manipulation
Unified Framework of Distributional Regret in Multi-Armed Bandits and Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
Driver-WM: A Driver-Centric Traffic-Conditioned Latent World Model for In-Cabin Dynamics RolloutWorld models → World models for autonomous driving
Think-Aloud Reshapes Automated Cognitive Model Discovery Beyond BehaviorLLM reasoning → Reasoning benchmarks and evaluation
Provable imitation learning for control of instability in partially-observed Vlasov--Poisson equationsOffline RL and imitation learning from data → Imitation learning methods
FlowDIS: Language-Guided Dichotomous Image Segmentation with Flow MatchingDiffusion models → Medical imaging and segmentation
SoK: Robustness in Large Language Models against Jailbreak AttacksLLM safety and security → Jailbreaking
Direct Product Flow Matching: Decoupling Radial and Angular Dynamics for Few-Shot AdaptationMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP +1
Piper: Efficient Large-Scale MoE Training via Resource Modeling and Pipelined Hybrid ParallelismMixture-of-experts → Efficient MoE training and inference systems
When Relations Break: Analyzing Relation Hallucination in Vision-Language Model Under Rotation and NoiseMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Preference-Based Self-Distillation: Beyond KL Matching via Reward RegularizationKnowledge distillation → On-policy and self-distillation for LLMs +1
Computer-Aided Design Generation by Cascaded Discrete Diffusion ModelDiffusion models → 3D, point cloud and scene generation
Local Intrinsic Dimension Unveils Hallucinations in Diffusion ModelsDiffusion models → Medical imaging and segmentation
Uno-Orchestra: Parsimonious Agent Routing via Selective DelegationLLM agents and tool use → General LLM agent frameworks and benchmarks
Misaligned by Reward: Socially Undesirable Preferences in LLMsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Agentic Vulnerability Reasoning on COTS BinariesLLM agents and tool use → Agent threats, guardrails and governance
Tailoring Scaffolding to Diagnostic Strategies: Theory-Informed LLM-Based AgentsLLM agents and tool use → Educational agents and math problem solving
Adaptivity Under Realizability Constraints: Comparing In-Context and Agentic LearningLLM agents and tool use → General LLM agent frameworks and benchmarks
You Snooze, You Lose: Automatic Safety Alignment Restoration through Neural Weight TranslationLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Self-Induced Outcome Potential: Turn-Level Credit Assignment for Agents without VerifiersLLM reasoning → Reinforcement learning for reasoning +1
Why Expert Alignment Is Hard: Evidence from Subjective EvaluationLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
EP-GRPO: Entropy-Progress Aligned Group Relative Policy Optimization with Implicit Process GuidanceLLM reasoning → Reinforcement learning for reasoning +1
Adaptive Inverted-Index Routing for Granular Mixtures-of-ExpertsMixture-of-experts → Expert routing strategies
UFAL-CUNI at SemEval-2026 Task 11: An Efficient Modular Neuro-symbolic Method for Syllogistic ReasoningLLM reasoning → Reasoning benchmarks and evaluation
Modular Reinforcement Learning For Cooperative SwarmsDeep reinforcement learning methods → RL for robot navigation and UAV control
Evolving Idea Graphs with Learnable Edits-and-Commits for Multi-Agent Scientific IdeationLLM agents and tool use → Web, search, and research agents
Reinforcement Learning for Compositional Generalization with Outcome-Level OptimizationDeep reinforcement learning methods → Deep RL algorithms and representations
Rethinking Local Learning: A Cheaper and Faster Recipe for LLM Post-TrainingLLM reasoning → Test-time compute and efficient thinking
Strat-Reasoner: Reinforcing Strategic Reasoning of LLMs in Multi-Agent GamesLLM reasoning → Reasoning for agents, planning and code +1
Storage Is Not Memory: A Retrieval-Centered Architecture for Agent RecallLLM agents and tool use → Agent memory systems
FairEnc: A Fair Vision-Language Model with Fair Vision and Text Encoders for Glaucoma DetectionMultimodal LLMs / vision-language models → Medical vision-language models
A Harmonic Mean Formulation of Average Reward Reinforcement Learning in SMDPsDeep reinforcement learning methods → RL theory and sample complexity
Uncertainty-Aware Exploratory Direct Preference Optimization for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs +1
VTAgent: Agentic Keyframe Anchoring for Evidence-Aware Video TextVQALLM agents and tool use → Multimodal and embodied agents
Agentic Repository Mining: A Multi-Task EvaluationLLM agents and tool use → Software engineering coding agents
StoryAlign: Evaluating and Training Reward Models for Story GenerationLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Concurrence of Symmetry Breaking and Nonlocality Phase Transitions in Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Tree-based Credit Assignment for Multi-Agent Memory SystemLLM agents and tool use → Agent memory systems

All 469