Papers

New on arXiv

Fri 24 Apr 2026

311 papers

EverydayGPT: Confidence-Gated Routing for Efficient and Safe Hybrid GPT-RAG Conversational QARetrieval-augmented generation → Conversational and speech RAG
Calibration Drift Under Reasoning: How Chain-of-Thought Budgets Induce Overconfidence in Large Language ModelsLLM reasoning → Test-time compute and efficient thinking
Multilingual Refusal Alignment for Safer Large Language ModelsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
Real-Time AttentionBender: Granular Interactive Network Bending of Video Diffusion TransformersDiffusion models → Video generation
Make Mechanistic Interpretability Auditable: A Call to Develop Guidelines via Continuous Collaborative ReviewingMechanistic interpretability → Circuits and mechanistic analysis
RAG4Outcome: A Retrieval-Augmented Multimodal Framework for Prognostic Prediction in Chronic OsteomyelitisRetrieval-augmented generation → Medical and biomedical RAG
From Demographics to Survey Anchors: Evaluating LLM Agents for Modeling Retirement AttitudesLLM agents and tool use → Social simulation and multi-agent debate
RL Token: Bootstrapping Online RL with Vision-Language-Action ModelsVision-language-action models → Reinforcement learning and post-training +1
Analytica: Soft Propositional Reasoning for Robust and Scalable LLM-Driven AnalysisLLM reasoning → Reasoning for agents, planning and code
ContextWeaver: Selective and Dependency-Structured Memory Construction for LLM AgentsLLM agents and tool use → Agent memory systems
Training a General Purpose Automated Red Teaming ModelLLM safety and security → Prompt injection and agent security
Don't Make the LLM Read the Graph: Make the Graph ThinkLLM reasoning → Reasoning for agents, planning and code
K-Score: Kalman Filter as a Principled Alternative to Reward Normalization in Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
A Decoupled Human-in-the-Loop System for Controlled Autonomy in Agentic WorkflowsLLM agents and tool use → General LLM agent frameworks and benchmarks
Shape of Memory: a Geometric Analysis of Machine Unlearning in Second-Order OptimizersLLM safety and security → Machine unlearning
Preserving Long-Tailed Expert Information in Mixture-of-Experts TuningMixture-of-experts → Expert routing strategies
GenAssets: Generating in-the-wild 3D Assets in Latent SpaceDiffusion models → Diffusion and flow policies for robotics
FormalScience: Scalable Human-in-the-Loop Autoformalisation of Science with Agentic Code Generation in LeanLLM reasoning → Mathematical reasoning +1
Vision-Language-Action in Robotics: A Survey of Datasets, Benchmarks, and Data EnginesVision-language-action models → Generalist VLA models and benchmarks
Learning from the Best: Smoothness-Driven Metrics for Data Quality in Imitation LearningOffline RL and imitation learning from data → Robot manipulation imitation learning
CheXmix: Unified Generative Pretraining for Vision Language Models in Medical ImagingMultimodal LLMs / vision-language models → Medical vision-language models
Uncertainty Quantification for LLM Function-CallingLLM agents and tool use → General LLM agent frameworks and benchmarks
Reward Models Are Secretly Value Functions: Temporally Coherent Reward ModelingLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Peer Identity Bias in Multi-Agent LLM Evaluation: An Empirical Study Using the TRUST Democratic Discourse Analysis PipelineLLM agents and tool use → Social simulation and multi-agent debate
VS-DDPM: Efficient Low-Cost Diffusion Model for Medical Modality TranslationDiffusion models → Medical imaging and segmentation
PExA: Parallel Exploration Agent for Complex Text-to-SQLLLM agents and tool use → Web, search, and research agents
Accelerating Frequency Domain Diffusion Models with Error-Feedback Event-Driven CachingDiffusion models → Diffusion foundations and sampling
Quantifying and Mitigating Self-Preference Bias of LLM JudgesLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
RouteGuard: Internal-Signal Detection of Skill Poisoning in LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
Can Multimodal Large Language Models Truly Understand Small Objects?Multimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Beyond Single-Agent Alignment: Preventing Context-Fragmented Violations in Multi-Agent SystemsLLM agents and tool use → Agent threats, guardrails and governance
How Do AI Agents Spend Your Money? Analyzing and Predicting Token Consumption in Agentic Coding TasksLLM agents and tool use → Agent memory systems
Agentic World Modeling: Foundations, Capabilities, Laws, and BeyondWorld models → World models for language agents +1
GCImOpt: Learning efficient goal-conditioned policies by imitating optimal trajectoriesOffline RL and imitation learning from data → Imitation learning methods +1
Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-ThoughtLLM reasoning → Prompting and chain-of-thought reasoning
Seeing the Whole Elephant: A Benchmark for Failure Attribution in LLM-based Multi-Agent SystemsLLM agents and tool use → Software engineering coding agents
4DLoG: Generative Modeling of Neurodegenerative Brain Anatomy with 4D Longitudinal Diffusion ModelDiffusion models → Medical imaging and segmentation
BERAG: Bayesian Ensemble Retrieval-Augmented Generation for Knowledge-based Visual Question AnsweringRetrieval-augmented generation → Multimodal RAG
Can QPP Choose the Right Query Variant? Evaluating Query Variant Selection for RAG PipelinesRetrieval-augmented generation → RAG robustness and adaptive retrieval
Structure-Guided Diffusion Model for EEG-Based Visual Cognition ReconstructionDiffusion models → Medical imaging and segmentation
Beyond Patient Invariance: Learning Cardiac Dynamics via Action-Conditioned JEPAsWorld models → Medical and surgical world models
Rethinking Math Reasoning Evaluation: A Robust LLM-as-a-Judge Framework Beyond Symbolic RigidityLLM reasoning → Mathematical reasoning
RedVLA: Physical Red Teaming for Vision-Language-Action ModelsVision-language-action models → Safety and adversarial attacks
Cross-Stage Coherence in Hierarchical Driving VQA: Explicit Baselines and Learned Gated Context ProjectorsEnd-to-end autonomous driving → VLMs for autonomous driving
SOLAR-RL: Semi-Online Long-horizon Assignment Reinforcement LearningMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +2
CGC: Compositional Grounded Contrast for Fine-Grained Multi-Image UnderstandingMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Towards Adaptive Continual Model Merging via Manifold-Aware Expert EvolutionMixture-of-experts → Mixture of LoRA experts
Superminds Test: Actively Evaluating Collective Intelligence of Agent Society via Probing AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
SSG: Logit-Balanced Vocabulary Partitioning for LLM WatermarkingLLM safety and security → Watermarking generated content and models
Automation-Exploit: A Multi-Agent LLM Framework for Adaptive Offensive Security with Digital Twin-Based Risk-Mitigated ExploitationLLM agents and tool use → Agent threats, guardrails and governance +1
SpaMEM: Benchmarking Dynamic Spatial Reasoning via Perception-Memory Integration in Embodied EnvironmentsLLM reasoning → Spatial and embodied reasoning +1
Efficient Diffusion Distillation via Embedding LossDiffusion models → Diffusion foundations and sampling
PoseFM: Relative Camera Pose Estimation Through Flow MatchingDiffusion models → 3D, point cloud and scene generation
Context-Fidelity Boosting: Enhancing Faithful Generation through Watermark-Inspired DecodingLLM safety and security → Watermarking generated content and models
Rethinking AI-Mediated Minority Support in Power-Imbalanced Group Decision-Making: From Anonymity To AuthenticityLLM reasoning → Reasoning for agents, planning and code
Guess-Verify-Refine: Data-Aware Top-K for Sparse-Attention Decoding on Blackwell via Temporal CorrelationEfficient LLM inference and serving → KV cache optimization
STEM: Structure-Tracing Evidence Mining for Knowledge Graphs-Driven Retrieval-Augmented GenerationLLM reasoning → Knowledge- and tool-augmented reasoning +1
DocPrune:Efficient Document Question Answering via Background, Question, and Comprehension-aware Token PruningMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Beyond Chain-of-Thought: Rewrite as a Universal Interface for Generative Multimodal EmbeddingsLLM reasoning → Multimodal reasoning +1
Multi-Agent Consensus as a Cognitive Bias Trigger in Human-AI InteractionLLM agents and tool use → Social simulation and multi-agent debate
Self-Correction as Feedback Control: Error Dynamics, Stability Thresholds, and Prompt Interventions in LLMsLLM agents and tool use → General LLM agent frameworks and benchmarks
Large Language Models Decide Early and Explain LaterLLM reasoning → Test-time compute and efficient thinking
Bridging the Long-Tail Gap: Robust Retrieval-Augmented Relation Completion via Multi-Stage Paraphrase InfusionRetrieval-augmented generation → RAG pipeline design and evaluation
Fast Neural-Network Approximation of Active Target Search Under UncertaintyLLM agents and tool use → Multimodal and embodied agents
OccDirector: Language-Guided Behavior and Interaction Generation in 4D Occupancy SpaceWorld models → World models for autonomous driving
CodeGraphVLP: Code-as-Planner Meets Semantic-Graph State for Non-Markovian Vision-Language-Action ModelsVision-language-action models → Reasoning and chain-of-thought in VLA
Preserve Support, Not Correspondence: Dynamic Routing for Offline Reinforcement LearningOffline RL and imitation learning from data → Offline RL algorithms
Towards Temporal Compositional Reasoning in Long-Form Sports VideosLLM reasoning → Multimodal reasoning
Breaking Watermarks in the Frequency Domain: A Modulated Diffusion Attack FrameworkLLM safety and security → Watermarking generated content and models
RAG-Reflect: Agentic Retrieval-Augmented Generation with Reflections for Comment-Driven Code Maintenance on Stack OverflowRetrieval-augmented generation → RAG for code +1
Multimodal Diffusion to Mutually Enhance Polarized Light and Low Resolution EBSD DataDiffusion models → Inverse problems and image restoration
UniSonate: A Unified Model for Speech, Music, and Sound Effect Generation with Text InstructionsDiffusion models → Audio, speech and music
Rethinking Semantic Collaborative Integration: Why Alignment Is Not EnoughLLM reasoning → LLM reasoning for recommendation
CharTide: Data-Centric Chart-to-Code Generation via Tri-Perspective Tuning and Inquiry-Driven EvolutionMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Behavioral Canaries: Auditing Private Retrieved Context Usage in RL Fine-TuningLLM agents and tool use → Agent threats, guardrails and governance
ReCast: Recasting Learning Signals for Reinforcement Learning in Generative RecommendationDeep reinforcement learning methods → RL training of language and agent models
Optimal sequential decision-making for error propagation mitigation in digital twinsDeep reinforcement learning methods → RL theory and sample complexity
Estimating Tail Risks in Language Model Output DistributionsLLM safety and security → Safety alignment, refusal and guardrails
Fine-Grained Analysis of Shared Syntactic Mechanisms in Language ModelsMechanistic interpretability → Knowledge and representations inside LLMs
Sum-of-Checks: Structured Reasoning for Surgical Safety with Large Vision-Language ModelsLLM reasoning → Clinical and biomedical applications +1
Reliable Self-Harm Risk Screening via Adaptive Multi-Agent LLM SystemsLLM agents and tool use → Medical and scientific domain agents
dWorldEval: Scalable Robotic Policy Evaluation via Discrete Diffusion World ModelWorld models → Embodied and robot world models +1
Global Convergence of Policy Gradient Methods for ReLU Controllers in Linear Quadratic RegulationDeep reinforcement learning methods → RL for games and combinatorial optimization
Sovereign Agentic Loops: Decoupling AI Reasoning from Execution in Real-World SystemsLLM reasoning → Reasoning for agents, planning and code +1
SHAPE: Unifying Safety, Helpfulness and Pedagogy for Educational LLMsLLM safety and security → Safety alignment, refusal and guardrails
Some Large Language Models Exhibit Consistent Risk Attitudes
Scaling Point-in-Time Language Models: Enabling Economic Evaluation of Embeddings
On the Identifiability of User Adaptation in Co-Adaptive Neural Interfaces
Ensembles of Large Language Models for Identifying EQ-5D Studies in PubMed Based on Their Abstracts
Exposing the Unsaid: Visualizing Hidden LLM Bias through Stochastic Path Aggregation
T2MM: An LLM Supported Architecture For Inquiry-Based Modeling
An LLM-Native Psychometric Instrument Reveals a Self-Report--Behavior Gap Across 25 Models
Smaller, Younger, and More Impactful: How AI-Assisted Writing Transforms Research Teams
LLM-assisted sentiment analysis for integrated computational and qualitative mixed methods education research: A case study of students' written reflection assignments
REC-CBM: Rubric-Aware Error-Correction Concept Bottleneck Models for Trustworthy Open-Ended Grading
Operationalizing Reconstructive Authority: Runtime Construction, Dependency Resolution, and Execution Gating in Autonomous Agent Systems
Practical Quantum CIM Empowerment via All-Domestic-Core Agentic Large Model
KT4EQG: Personalized Exercise Question Generation via Knowledge Tracing
Toeplitz MLP Mixers are Low Complexity, Information-Rich Sequence Models
Fast and Effective Redistricting Optimization via Composite-Move Tabu Search

All 311