Papers

New on arXiv

Wed 22 Apr 2026

417 papers

C2MOE: Consistency and Complementarity-guided Mixture of Experts for Incomplete Multimodal Emotion LearningMixture-of-experts → Multimodal MoE models
Causal-Audit: Explicit and Auditable Graph-based Reasoning via Target-Aware Causal Chain ConstructionLLM reasoning → Causal reasoning in LLMs
Compatibility-Aware Dynamic Fine-Tuning for Large Language ModelsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Dual-Stance Evaluation of Sycophancy: The Structure of Agreement and the Limits of InterventionMechanistic interpretability → Activation steering methods
One Jailbreak, Many Tongues: Learning Language-Insensitive Intention Representations for Multilingual Jailbreak DetectionLLM safety and security → Jailbreaking
To Intervene or Not: Guiding Inference-time Alignment with Probabilistic Model BlendingLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Quantum Frog: Emergent Cooperation and Difficulty Scaling in a Quantized-Time Cooperative GameDeep reinforcement learning methods → RL for games and combinatorial optimization
On the Role of Strain and Vorticity in Numerical Integration Error for Flow MatchingDiffusion models → Flow matching, time series and science
Breaking the Illusion: When Positive Meets Negative in Multimodal DecodingMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
LKV: End-to-End Learning of Head-wise Budgets and Token Selection for LLM KV Cache EvictionEfficient LLM inference and serving → KV cache optimization
RateQuant: Optimal Mixed-Precision KV Cache Quantization via Rate-Distortion TheoryEfficient LLM inference and serving → KV cache quantization
EvoJail: Evolutionary Diverse Jailbreak Prompt Generation for Large Language ModelsLLM safety and security → Jailbreaking
SwarmDrive: Semantic V2V Coordination for Latency-Constrained Cooperative Autonomous DrivingEnd-to-end autonomous driving → Cooperative and V2X driving
EgoDyn-Bench: Evaluating Ego-Motion Understanding in Vision-Centric Foundation Models for Autonomous DrivingLLM reasoning → Chain-of-thought reasoning for driving +2
R3AG: Retriever Routing for Retrieval-Augmented GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
Adaptive Instruction Composition for Automated LLM Red-TeamingLLM safety and security → Prompt injection and agent security
Dialect vs Demographics: Quantifying LLM Bias from Implicit Linguistic Signals vs. Explicit User ProfilesLLM safety and security → Safety alignment, refusal and guardrails
WFM: 3D Wavelet Flow Matching for Ultrafast Multi-Modal MRI SynthesisDiffusion models → Medical imaging and segmentation
Enhancing Science Classroom Discourse Analysis through Joint Multi-Task Learning for Reasoning-Component ClassificationLLM reasoning → Mathematical reasoning
Beyond Pixels: Introspective and Interactive Grounding for Visualization AgentsMultimodal LLMs / vision-language models → Chart understanding and anomaly detection +1
Cross-Session Threats in AI Agents: Benchmark, Evaluation, and AlgorithmsLLM agents and tool use → Agent threats, guardrails and governance +1
Self-Predictive Representation for Autonomous UAV Object-Goal NavigationDeep reinforcement learning methods → RL for robot navigation and UAV control
Leveraging Multimodal LLMs for Built Environment and Housing Attribute Assessment from Street-View ImageryMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Weighting What Matters: Boosting Sample Efficiency in Medical Report Generation via Token ReweightingMultimodal LLMs / vision-language models → Medical vision-language models
Foveated Reasoning: Stateful, Action-based Visual Focusing for Vision-Language ModelsLLM reasoning → Multimodal reasoning +1
Distributed Generative Inference of LLM at Internet Scales with Multi-Dimensional Communication OptimizationEfficient LLM inference and serving → Speculative decoding
Optimizing Diffusion Priors in Image Reconstruction from a Single ObservationDiffusion models → Inverse problems and image restoration
InVitroVision: a Multi-Modal AI Model for Automated Description of Embryo Development using Natural LanguageMultimodal LLMs / vision-language models → Medical vision-language models
TRACES: Tagging Reasoning Steps for Adaptive Cost-Efficient Early-StoppingLLM reasoning → Test-time compute and efficient thinking
Strategic Polysemy in AI Discourse: A Philosophical Analysis of Language, Hype, and PowerLLM reasoning → Reasoning benchmarks and evaluation
Projected Gradient Unlearning for Text-to-Image Diffusion Models: Defending Against Concept Revival AttacksDiffusion models → Robustness, privacy and adversarial aspects +1
Who Defines Fairness? Target-Based Prompting for Demographic Representation in Generative ModelsDiffusion models → Image generation, text-to-image and editing
A Systematic Review and Taxonomy of Reinforcement Learning-Model Predictive Control Integration for Linear SystemsDeep reinforcement learning methods → RL for autonomous driving and traffic
Adaptive Test-Time Compute Allocation with Evolving In-Context DemonstrationsLLM reasoning → Test-time compute and efficient thinking
Open-H-Embodiment: A Large-Scale Dataset for Enabling Foundation Models in Medical RoboticsWorld models → Medical and surgical world models
Linear Image Generation by Synthesizing Exposure BracketsDiffusion models → Image generation, text-to-image and editing
AFRILANGTUTOR: Advancing Language Tutoring and Culture Education in Low-Resource Languages with Large Language ModelsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Breaking MCP with Function Hijacking Attacks: Novel Threats for Function Calling and Agentic ModelsLLM agents and tool use → Agent threats, guardrails and governance +1
Co-Evolving LLM Decision and Skill Bank Agents for Long-Horizon TasksLLM agents and tool use → Reinforcement learning for LLM agents
Thinking Like a Botanist: Challenging Multimodal Language Models with Intent-Driven Chain-of-InquiryLLM reasoning → Clinical and biomedical applications +1
Can Virtual Agents Care? Designing an Empathetic and Personalized LLM-Driven Conversational AgentLLM agents and tool use → Medical and scientific domain agents
Breaking Bad: Interpretability-Based Safety Audits of State-of-the-Art LLMsLLM safety and security → Safety alignment, refusal and guardrails
HARBOR: Automated Harness OptimizationLLM agents and tool use → Software engineering coding agents
AttentionBender: Manipulating Cross-Attention in Video Diffusion Transformers as a Creative ProbeDiffusion models → Video generation
IRIS: Interpolative R'enyi Iterative Self-play for Large Language Model Fine-TuningSynthetic data and self-improvement for LLMs → Self-rewarding alignment and self-training
Adaptive Defense Orchestration for RAG: A Sentinel-Strategist Architecture against Multi-Vector AttacksRetrieval-augmented generation → Privacy and security of RAG +1
SafeRedirect: Defeating Internal Safety Collapse via Task-Completion Redirection in Frontier LLMsLLM safety and security → Safety alignment, refusal and guardrails
Learning Reasoning World Models for Parallel CodeLLM reasoning → Reasoning for agents, planning and code
Simplified Sparse Attention via Gist TokensEfficient LLM inference and serving → KV cache optimization
The Path Not Taken: Duality in Reasoning about Program ExecutionLLM reasoning → Reasoning for agents, planning and code
Omission Constraints Decay While Commission Constraints Persist in Long-Context LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
PokeVLA: Empowering Pocket-Sized Vision-Language-Action Model with Comprehensive World Knowledge GuidanceVision-language-action models → 3D and spatial-aware VLA
AVISE: Framework for Evaluating the Security of AI SystemsLLM safety and security → Jailbreaking
ParetoSlider: Diffusion Models Post-Training for Continuous Reward ControlDiffusion models → Diffusion and flow policies for robotics +2
Diagnosing CFG Interpretation in LLMsLLM agents and tool use → General LLM agent frameworks and benchmarks
OMIBench: Benchmarking Olympiad-Level Multi-Image Reasoning in Large Vision-Language ModelLLM reasoning → Multimodal reasoning +1
Synthesizing Multi-Agent Harnesses for Vulnerability DiscoveryLLM agents and tool use → Software engineering coding agents
LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language ModelDiffusion models → Image generation, text-to-image and editing
SWE-chat: Coding Agent Interactions From Real Users in the WildLLM agents and tool use → Software engineering coding agents
Coverage, Not Averages: Semantic Stratification for Trustworthy Retrieval EvaluationRetrieval-augmented generation → RAG robustness and adaptive retrieval
V-tableR1: Process-Supervised Multimodal Table Reasoning with Critic-Guided Policy OptimizationLLM reasoning → Multimodal reasoning +1
Where and What: Reasoning Dynamic and Implicit Preferences in Situated Conversational RecommendationLLM reasoning → LLM reasoning for recommendation +1
RespondeoQA: a Benchmark for Bilingual Latin-English Question AnsweringLLM reasoning → Reasoning benchmarks and evaluation
Learning from the Near Future: Temporal Self-Distillation for RLVRKnowledge distillation → On-policy and self-distillation for LLMs
Render-in-the-Loop: Vector Graphics Generation via Visual Self-FeedbackMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Supplement Generation Training for Enhancing Agentic Task PerformanceLLM agents and tool use → General LLM agent frameworks and benchmarks
Tokenised Flow Matching for Hierarchical Simulation Based InferenceDiffusion models → Flow matching, time series and science
GeoRelight: Learning Joint Geometrical Relighting and Reconstruction with Flexible Multi-Modal Diffusion TransformersDiffusion models → 3D, point cloud and scene generation
SSL-R1: Self-Supervised Visual Reinforcement Post-Training for Multimodal Large Language ModelsLLM reasoning → Multimodal reasoning +1
R-CoV: Region-Aware Chain-of-Verification for Alleviating Object Hallucinations in LVLMsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
MGDA-Decoupled: Geometry-Aware Multi-Objective Optimisation for DPO-based LLM AlignmentLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
ORPHEAS: A Cross-Lingual Greek-English Embedding Model for Retrieval-Augmented GenerationRetrieval-augmented generation → Generation and context integration
The Expense of Seeing: Attaining Trustworthy Multimodal Reasoning Within the Monolithic ParadigmLLM reasoning → Multimodal reasoning +1
GRPO-VPS: Enhancing Group Relative Policy Optimization with Verifiable Process Supervision for Effective ReasoningLLM reasoning → Reinforcement learning for reasoning +1
Cooperative Profiles Predict Multi-Agent LLM Team Performance in AI for Science WorkflowsLLM reasoning → Reasoning for agents, planning and code +1
CHORUS: An Agentic Framework for Generating Realistic Deliberation DataLLM agents and tool use → Social simulation and multi-agent debate
Occupancy Reward Shaping: Improving Credit Assignment for Offline Goal-Conditioned Reinforcement LearningWorld models → Latent dynamics models and model-based RL +1
Beyond ZOH: Advanced Discretization Strategies for Vision MambaState-space sequence models (S4, Mamba) → Vision Mamba backbones
Self-Guided Plan Extraction for Instruction-Following Tasks with Goal-Conditional Reinforcement LearningDeep reinforcement learning methods → RL training of language and agent models
Self-Aware Vector Embeddings for Retrieval-Augmented Generation: A Neuroscience-Inspired Framework for Temporal, Confidence-Weighted, and Relational KnowledgeRetrieval-augmented generation → Generation and context integration
Physics-Informed Conditional Diffusion for Motion-Robust Retinal Temporal Laser Speckle Contrast ImagingDiffusion models → Medical imaging and segmentation
Trust, Lies, and Long Memories: Emergent Social Dynamics and Reputation in Multi-Round Avalon with LLM AgentsLLM agents and tool use → Social simulation and multi-agent debate
Exploring Spatial Intelligence from a Generative PerspectiveMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Where Reasoning Breaks: Logic-Aware Path Selection by Controlling Logical Connectives in LLMs Reasoning ChainsLLM reasoning → Test-time compute and efficient thinking
LLM StructCore: Schema-Guided Reasoning Condensation and Deterministic CompilationLLM reasoning → Clinical and biomedical applications
LayerTracer: A Joint Task-Particle and Vulnerable-Layer Analysis framework for Arbitrary Large Language Model ArchitecturesState-space sequence models (S4, Mamba) → Core SSM architectures for language
On Bayesian Softmax-Gated Mixture-of-Experts ModelsMixture-of-experts → MoE theory and gating functions
Enhancing Research Idea Generation through Combinatorial Innovation and Multi-Agent Iterative Search StrategiesLLM agents and tool use → Web, search, and research agents
Evian: Towards Explainable Visual Instruction-tuning Data AuditingMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
FASER: Fine-Grained Phase Management for Speculative Decoding in Dynamic LLM ServingEfficient LLM inference and serving → Speculative decoding
Efficient Test-Time Inference via Deterministic Exploration of Truncated Decoding TreesLLM reasoning → Test-time compute and efficient thinking
Near-Field Wideband Channel Estimation for XL-MIMO Systems via Denoising Diffusion ModelDiffusion models → Diffusion for compression and wireless
ProMMSearchAgent: A Generalizable Multimodal Search Agent Trained with Process-Oriented RewardsLLM reasoning → Multimodal reasoning +1
Video-ToC: Video Tree-of-Cue ReasoningLLM reasoning → Multimodal reasoning
Temporal Difference Calibration in Sequential Tasks: Application to Vision-Language-Action ModelsVision-language-action models → Failure recovery and robustness
DynamicRad: Content-Adaptive Sparse Attention for Long Video DiffusionDiffusion models → Video generation
Mechanistic Interpretability Tool for AI Weather ModelsMechanistic interpretability → Interpreting time-series and weather models
CCTVBench: Contrastive Consistency Traffic VideoQA Benchmark for Multimodal LLMsMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection
HaS: Accelerating RAG through Homology-Aware Speculative RetrievalRetrieval-augmented generation → RAG pipeline design and evaluation
MedSkillAudit: A Domain-Specific Audit Framework for Medical Research Agent SkillsLLM agents and tool use → Web, search, and research agents

All 417