Papers

New on arXiv

Tue 14 Apr 2026

474 papers

Rater State Bias in RLHF Preference Data: An Audit FrameworkLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Epidemiology of Model Collapse: Modeling Synthetic Data Contamination via Bilayer SIR DynamicsSynthetic data and self-improvement for LLMs → Model collapse from synthetic data
SENSE: Semantic Embedding Navigation with Soft-gated Evaluation for Retrieval-based Speculative DecodingEfficient LLM inference and serving → Speculative decoding
CSRP: Chain-of-Thought Reasoning for Chinese Text Correction via Reinforcement Learning with Efficiency-Aware RewardsLLM reasoning → Reasoning benchmarks and evaluation
ClusterRAG: Cluster-Based Collaborative Filtering for Personalized Retrieval-Augmented GenerationRetrieval-augmented generation → RAG pipeline design and evaluation
Efficient Handwriting-Based Alzheimer,s Disease Diagnosis Using a Low-Rank Mixture of Experts Deep Learning FrameworkMixture-of-experts → Medical and biomedical MoE
Validity-Calibrated Reasoning DistillationLLM reasoning → Prompting and chain-of-thought reasoning +1
Balanced Aggregation: Understanding and Fixing Aggregation Bias in GRPOLLM reasoning → Reinforcement learning for reasoning
RetentiveKV: State-Space Memory for Uncertainty-Aware Multimodal KV Cache EvictionMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
MARS: Efficient, Adaptive Co-Scheduling for Heterogeneous Agentic SystemsLLM agents and tool use → Software engineering coding agents
FreqFormer: Hierarchical Frequency-Domain Attention with Adaptive Spectral Routing for Long-Sequence Video Diffusion TransformersDiffusion models → Video generation
Agent-GWO: Collaborative Agents for Dynamic Prompt Optimization in Large Language ModelsLLM reasoning → Reasoning for agents, planning and code +1
Medical thinking with multiple imagesLLM reasoning → Clinical and biomedical applications
From Handwriting to Structured Data: Benchmarking AI Digitisation of Handwritten FormsMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Topology-Aware Layer Pruning for Large Vision-Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
HQA-VLAttack: Towards High Quality Adversarial Attack on Vision-Language Pre-Trained ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Dive into Claude Code: The Design Space of Today's and Future AI Agent SystemsLLM agents and tool use → Software engineering coding agents
FRESCO: Benchmarking and Optimizing Re-rankers for Evolving Semantic Conflict in Retrieval-Augmented GenerationRetrieval-augmented generation → Generation and context integration
TRACE: A Conversational Framework for Sustainable Tourism Recommendation with Agentic Counterfactual ExplanationsLLM agents and tool use → Agentic recommender systems
Adaptive Query Routing: A Tier-Based Framework for Hybrid Retrieval Across Financial, Legal, and Medical DocumentsRetrieval-augmented generation → Legal and financial RAG
Knowledge Graph RAG: Agentic Crawling and Graph Construction in Enterprise DocumentsRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG +1
Peer-Predictive Self-Training for Language Model ReasoningLLM reasoning → Test-time compute and efficient thinking +1
Finetuning-Free Diffusion Model with Adaptive Constraint Guidance for Inorganic Crystal Structure GenerationDiffusion models → Flow matching, time series and science
When Less Latent Leads to Better Relay: Information-Preserving Compression for Latent Multi-Agent LLM CollaborationLLM agents and tool use → General LLM agent frameworks and benchmarks
AgentSPEX: An Agent SPecification and EXecution LanguageLLM agents and tool use → Software engineering coding agents
Multi-Agent Object Detection Framework Based on Raspberry Pi YOLO Detector and Slack-Ollama Natural Language InterfaceLLM agents and tool use → Multimodal and embodied agents
Why MLLMs Struggle to Determine Object OrientationsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
WebXSkill: Skill Learning for Autonomous Web AgentsLLM agents and tool use → Web, search, and research agents
Concrete Jungle: Towards Concreteness Paved Contrastive Negative Mining for Compositional UnderstandingMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Bias at the End of the ScoreLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Can Cross-Layer Transcoders Replace Vision Transformer Activations? An Interpretable Perspective on VisionMechanistic interpretability → Interpreting vision and multimodal models
Can Agents Secure Hardware? Evaluating Agentic LLM-Driven Obfuscation for IP ProtectionLLM agents and tool use → Agent threats, guardrails and governance
Indexing Multimodal Language Models for Large-scale Image RetrievalLLM reasoning → Multimodal reasoning +1
On the Creativity of AI AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
SemiFA: An Agentic Multi-Modal Framework for Autonomous Semiconductor Failure Analysis Report GenerationLLM agents and tool use → Software engineering coding agents
Numerical Instability and Chaos: Quantifying the Unpredictability of Large Language ModelsLLM agents and tool use → General LLM agent frameworks and benchmarks
Unleashing Implicit Rewards: Prefix-Value Learning for Distribution-Level OptimizationLLM reasoning → Reinforcement learning for reasoning +1
Synthesis and Deployment of Maximal Robust Control Barrier Functions through Adversarial Reinforcement LearningDeep reinforcement learning methods → Safe and constrained RL
SciFi: A Safe, Lightweight, User-Friendly, and Fully Autonomous Agentic AI Workflow for Scientific ApplicationsLLM agents and tool use → Web, search, and research agents
Pareto-Optimal Offline Reinforcement Learning via Smooth Tchebycheff ScalarizationOffline RL and imitation learning from data → Offline RL algorithms
Exploration and Exploitation Errors Are Measurable for Language Model AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Automated co-design of high-performance thermodynamic cycles via graph-based hierarchical reinforcement learningDeep reinforcement learning methods → RL for energy, networks and operations
SceneCritic: A Symbolic Evaluator for 3D Indoor Scene SynthesisMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Generative Refinement Networks for Visual SynthesisDiffusion models → Diffusion foundations and sampling
Visual Preference Optimization with Rubric RewardsLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Representation geometry shapes task performance in vision-language modeling for CT enterographyMultimodal LLMs / vision-language models → Medical vision-language models
Toward Autonomous Long-Horizon Engineering for ML ResearchLLM agents and tool use → Web, search, and research agents
Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and RecipeLLM reasoning → Prompting and chain-of-thought reasoning +1
Lightning OPD: Efficient Post-Training for Large Reasoning Models with Offline On-Policy DistillationLLM reasoning → Prompting and chain-of-thought reasoning +1
Agentic Discovery with Active Hypothesis Exploration for Visual RecognitionLLM agents and tool use → Multimodal and embodied agents
Causal Diffusion Models for Counterfactual Outcome Distributions in Longitudinal DataDiffusion models → Diffusion foundations and sampling
Accelerating Speculative Decoding with Block Diffusion Draft TreesEfficient LLM inference and serving → Speculative decoding
Parallax: Why AI Agents That Think Must Never ActLLM safety and security → Prompt injection and agent security
GlotOCR Bench: OCR Models Still Struggle Beyond a Handful of Unicode ScriptsMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Cycle-Consistent Search: Question Reconstructability as a Proxy Reward for Search Agent TrainingDeep reinforcement learning methods → RL training of language and agent models
Boosting Visual Instruction Tuning with Self-Supervised GuidanceLLM reasoning → Multimodal reasoning +1
Drawing on Memory: Dual-Trace Encoding Improves Cross-Session Recall in LLM AgentsLLM reasoning → Reasoning for agents, planning and code +1
Tree Learning: A Multi-Skill Continual Learning Framework for Humanoid RobotsDeep reinforcement learning methods → RL for robot locomotion and manipulation
Don't Show Pixels, Show Cues: Unlocking Visual Tool Reasoning in Language Models via Perception ProgramsLLM reasoning → Multimodal reasoning +1
Towards Long-horizon Agentic Multimodal SearchLLM agents and tool use → Multimodal and embodied agents
Evaluating LLMs Code Reasoning Under Real-World ContextLLM reasoning → Reasoning for agents, planning and code
LIFE -- an energy efficient advanced continual learning agentic AI framework for frontier systemsLLM agents and tool use → General LLM agent frameworks and benchmarks
QuarkMedSearch: A Long-Horizon Deep Search Agent for Exploring Medical IntelligenceLLM agents and tool use → Medical and scientific domain agents
Challenging Vision-Language Models with Physically Deployable Multimodal Semantic Lighting AttacksMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
RePAIR: Interactive Machine Unlearning through Prompt-Aware Model RepairLLM safety and security → Machine unlearning
Understanding and Improving Continuous Adversarial Training for LLMs via In-context Learning TheoryLLM safety and security → Jailbreaking
DPC-VQA: Decoupling Quality Perception and Residual Calibration for Video Quality AssessmentMultimodal LLMs / vision-language models → Image editing and quality assessment
DocSeeker: Structured Visual Reasoning with Evidence Grounding for Long Document UnderstandingLLM reasoning → Multimodal reasoning +1
EvoSpark: Endogenous Interactive Agent Societies for Unified Long-Horizon Narrative EvolutionLLM agents and tool use → Multimodal and embodied agents
CLASP: Class-Adaptive Layer Fusion and Dual-Stage Pruning for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
NaviRAG: Towards Active Knowledge Navigation for Retrieval-Augmented GenerationRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
ARGOS: Who, Where, and When in Agentic Multi-Camera Person SearchLLM agents and tool use → Multimodal and embodied agents
Generating Effective CoT Traces for Mitigating Causal HallucinationLLM reasoning → Hallucination detection and mitigation
Evaluating Differential Privacy Against Membership Inference in Federated Learning: Insights from the NIST Genomics Red Team ChallengeLLM safety and security → Federated unlearning and poisoning
Token-Level Policy Optimization: Linking Group-Level Rewards to Token-Level Aggregation via Sequence-Level LikelihoodLLM reasoning → Reinforcement learning for reasoning
AffectAgent: Collaborative Multi-Agent Reasoning for Retrieval-Augmented Multimodal Emotion RecognitionLLM reasoning → Audio-language model reasoning +2
Transferable Expertise for Autonomous Agents via Real-World Case-Based LearningLLM agents and tool use → Agent skill learning and evolution
BID-LoRA: A Parameter-Efficient Framework for Continual Learning and UnlearningLLM safety and security → Machine unlearning
Brain-DiT: A Universal Multi-state fMRI Foundation Model with Metadata-Conditioned PretrainingDiffusion models → Medical imaging and segmentation
A hierarchical spatial-aware algorithm with efficient reinforcement learning for human-robot task planning and allocation in productionDeep reinforcement learning methods → RL for energy, networks and operations
OFA-Diffusion Compression: Compressing Diffusion Model in One-Shot MannerDiffusion models → Diffusion for compression and wireless
Safe reinforcement learning with online filtering for fatigue-predictive human-robot task planning and allocation in productionDeep reinforcement learning methods → RL for energy, networks and operations
From Imitation to Discrimination: Progressive Curriculum Learning for Robust Web NavigationLLM agents and tool use → Web, search, and research agents
Do VLMs Truly "Read" Candlesticks? A Multi-Scale Benchmark for Visual Stock Price ForecastingMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
FeaXDrive: Feasibility-aware Trajectory-Centric Diffusion Planning for End-to-End Autonomous DrivingEnd-to-end autonomous driving → Imitation-learning end-to-end planners
PromptEcho: Annotation-Free Reward from Vision-Language Models for Text-to-Image Reinforcement LearningMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
Learning Chain Of Thoughts Prompts for Predicting Entities, Relations, and even Literals on Knowledge GraphsLLM reasoning → Prompting and chain-of-thought reasoning
Adaptive Test-Time Scaling for Zero-Shot Respiratory Audio ClassificationLLM reasoning → Audio-language model reasoning
Contextual Multi-Task Reinforcement Learning for Autonomous Reef MonitoringDeep reinforcement learning methods → RL for robot navigation and UAV control
Calibration-Aware Policy Optimization for Reasoning LLMsLLM reasoning → Reinforcement learning for reasoning
GeoAlign: Geometric Feature Realignment for MLLM Spatial ReasoningLLM reasoning → Spatial and embodied reasoning +1
A Comparison of Reinforcement Learning and Optimal Control Methods for Path PlanningDeep reinforcement learning methods → RL for robot navigation and UAV control
KnowRL: Boosting LLM Reasoning via Reinforcement Learning with Minimal-Sufficient Knowledge GuidanceLLM reasoning → Reinforcement learning for reasoning
SOAR: Self-Correction for Optimal Alignment and Refinement in Diffusion ModelsDiffusion models → Diffusion and flow policies for robotics +1
Every Picture Tells a Dangerous Story: Memory-Augmented Multi-Agent Jailbreak Attacks on VLMsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +2
Transforming External Knowledge into Triplets for Enhanced Retrieval in RAG of LLMsRetrieval-augmented generation → RAG pipeline design and evaluation
StructDiff: A Structure-Preserving and Spatially Controllable Diffusion Model for Single-Image GenerationDiffusion models → Image generation, text-to-image and editing
Cross-Modal Knowledge Distillation for PET-Free Amyloid-Beta Detection from MRIKnowledge distillation → Distillation for medical imaging
Cross-Attentive Multiview Fusion of Vision-Language EmbeddingsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
DeepSeek Robustness Against Semantic-Character Dual-Space Mutated Prompt InjectionLLM safety and security → Prompt injection and agent security

All 474