Papers

New on arXiv

Wed 19 Feb 2025

363 papers

Towards Adaptive Memory-Based Optimization for Enhanced Retrieval-Augmented GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
Think-Then-React: Towards Unconstrained Human Action-to-Reaction GenerationLLM reasoning → Spatial and embodied reasoning
Robust Optimization with Diffusion Models for Green SecurityDiffusion models → Diffusion and flow policies for robotics
Theoretical Physics Benchmark (TPBench) -- a Dataset and Study of AI Reasoning Capabilities in Theoretical PhysicsLLM reasoning → Spatial and embodied reasoning
InsightVision: A Comprehensive, Multi-Level Chinese-based Benchmark for Evaluating Implicit Visual Semantics in Large Vision Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Zero-Shot Commonsense Validation and Reasoning with Large Language Models: An Evaluation on SemEval-2020 Task 4 DatasetLLM reasoning → Reasoning benchmarks and evaluation
Black Sheep in the Herd: Playing with Spuriously Correlated Attributes for Vision-Language RecognitionMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
A Mousetrap: Fooling Large Reasoning Models for Jailbreak with Chain of Iterative ChaosLLM reasoning → Reasoning for agents, planning and code +1
FragFM: Hierarchical Framework for Efficient Molecule Generation via Fragment-Level Discrete Flow MatchingDiffusion models → Flow matching, time series and science
FairKV: Balancing Per-Head KV Cache for Fast Multi-GPU InferenceEfficient LLM inference and serving → KV cache optimization
CODEPROMPTZIP: Code-specific Prompt Compression for Retrieval-Augmented Generation in Coding Tasks with LMsRetrieval-augmented generation → RAG for code
SIFT: Grounding LLM Reasoning in Contexts via StickersLLM reasoning → Test-time compute and efficient thinking
The Canary's Echo: Auditing Privacy Risks of LLM-Generated Synthetic TextSynthetic data and self-improvement for LLMs → Privacy-preserving tabular data synthesis
Sce2DriveX: A Generalized MLLM Framework for Scene-to-Drive LearningEnd-to-end autonomous driving → VLMs for autonomous driving
CAPability: A Comprehensive Visual Caption Benchmark for Evaluating Both Correctness and ThoroughnessMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
SegSub: Evaluating Robustness to Knowledge Conflicts and Hallucinations in Vision-Language ModelsLLM reasoning → Hallucination detection and mitigation +1
Giving AI Personalities Leads to More Human-Like ReasoningLLM reasoning → Reasoning benchmarks and evaluation
PitVQA++: Vector Matrix-Low-Rank Adaptation for Open-Ended Visual Question Answering in Pituitary SurgeryMultimodal LLMs / vision-language models → Medical vision-language models
Efficient and Optimal Policy Gradient Algorithm for Corrupted Multi-armed BanditsDeep reinforcement learning methods → RL theory and sample complexity
UM_FHS at TREC 2024 PLABA: Exploration of Fine-tuning and AI agent approach for plain language adaptations of biomedical textLLM agents and tool use → Medical and scientific domain agents
A hierarchical resource-efficient deep policy gradient method for continuous-time optimal control problemsDeep reinforcement learning methods → RL theory and sample complexity
ModSkill: Physical Character Skill ModularizationOffline RL and imitation learning from data → Robot manipulation imitation learning
Self-Regularization with Sparse Autoencoders for Controllable LLM-based ClassificationMechanistic interpretability → SAE features and their applications
An Empirical Risk Minimization Approach for Offline Inverse RL and Dynamic Discrete Choice ModelOffline RL and imitation learning from data → Inverse reinforcement learning
Modular Prompt Learning Improves Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Understanding SGD with Exponential Moving Average: A Case Study in Linear RegressionDiffusion models → Diffusion foundations and sampling
Object-centric Binding in Contrastive Language-Image PretrainingMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Towards Context-Robust LLMs: A Gated Representation Fine-tuning ApproachRetrieval-augmented generation → RAG robustness and adaptive retrieval
MambaLiteSR: Image Super-Resolution with Low-Rank Mamba using Knowledge DistillationState-space sequence models (S4, Mamba) → Vision Mamba backbones +1
Navigating Semantic Relations: Challenges for Language Models in Abstract Common-Sense ReasoningLLM reasoning → Reasoning benchmarks and evaluation
Are Rules Meant to be Broken? Understanding Multilingual Moral Reasoning as a Computational Pipeline with UniMoralLLM reasoning → Reasoning benchmarks and evaluation
Investigating Non-Transitivity in LLM-as-a-JudgeLLM agents and tool use → General LLM agent frameworks and benchmarks
DiffExp: Efficient Exploration in Reward Fine-tuning for Text-to-Image Diffusion ModelsDiffusion models → Diffusion and flow policies for robotics
RocketKV: Accelerating Long-Context LLM Inference via Two-Stage KV Cache CompressionEfficient LLM inference and serving → KV cache optimization
Diversity-driven Data Selection for Language Model Tuning through Sparse AutoencoderMechanistic interpretability → SAE features and their applications
Enhancing Cognition and Explainability of Multimodal Foundation Models with Self-Synthesized DataMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Dynamic Activation with Knowledge Distillation for Energy-Efficient Spiking NN EnsemblesKnowledge distillation → Distillation for spiking neural networks
d-Sketch: Improving Visual Fidelity of Sketch-to-Image Translation with Pretrained Latent Diffusion Models without RetrainingDiffusion models → Image generation, text-to-image and editing
Im2SurfTex: Surface Texture Generation via Neural Backprojection of Multi-View ImagesDiffusion models → 3D, point cloud and scene generation
Human-Artificial Interaction in the Age of Agentic AI: A System-Theoretical ApproachLLM agents and tool use → General LLM agent frameworks and benchmarks
SigStyle: Signature Style Transfer via Personalized Text-to-Image ModelsDiffusion models → Image generation, text-to-image and editing
Beyond Single-Value Metrics: Evaluating and Enhancing LLM Unlearning with Cognitive DiagnosisLLM safety and security → Machine unlearning
FantasyID: Face Knowledge Enhanced ID-Preserving Video GenerationDiffusion models → Image generation, text-to-image and editing
Generative Detail Enhancement for Physically Based MaterialsDiffusion models → 3D, point cloud and scene generation
Remote Sensing Semantic Segmentation Quality Assessment based on Vision Language ModelMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
SelfAge: Personalized Facial Age Transformation Using Self-reference ImagesDiffusion models → Image generation, text-to-image and editing
Autellix: An Efficient Serving Engine for LLM Agents as General ProgramsLLM agents and tool use → Software engineering coding agents +1
Is That Your Final Answer? Test-Time Scaling Improves Selective Question AnsweringLLM reasoning → Test-time compute and efficient thinking
Supervising the search process produces reliable and generalizable information-seeking agentsLLM agents and tool use → Web, search, and research agents
Why Safeguarded Ships Run Aground? Aligned Large Language Models' Safety Mechanisms Tend to Be Anchored in The Template RegionLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
AdaptiveStep: Automatically Dividing Reasoning Step through Model ConfidenceLLM reasoning → Reinforcement learning for reasoning +1
A Chain-of-Thought Subspace Meta-Learning for Few-shot Image Captioning with Large Vision and Language ModelsLLM reasoning → Multimodal reasoning
Image compositing is all you need for data augmentationDiffusion models → Image generation, text-to-image and editing
Symmetrical Visual Contrastive Optimization: Aligning Vision-Language Models with Minimal Contrastive ImagesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Beyond Single Frames: Can LMMs Comprehend Temporal and Contextual Narratives in Image Sequences?Multimodal LLMs / vision-language models → MLLM training and visual instruction tuning
LongPO: Long Context Self-Evolution of Large Language Models through Short-to-Long Preference OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Playing Hex and Counter Wargames using Reinforcement Learning and Recurrent Neural NetworksDeep reinforcement learning methods → RL for games and combinatorial optimization
TESS 2: A Large-Scale Generalist Diffusion Language ModelDiffusion models → Discrete and text diffusion
How Do LLMs Perform Two-Hop Reasoning in Context?LLM reasoning → Knowledge- and tool-augmented reasoning
Optimistically Optimistic Exploration for Provably Efficient Infinite-Horizon Reinforcement and Imitation LearningOffline RL and imitation learning from data → Offline RL algorithms +1
DataSciBench: An LLM Agent Benchmark for Data ScienceLLM agents and tool use → Web, search, and research agents
DH-RAG: A Dynamic Historical Context-Powered Retrieval-Augmented Generation Method for Multi-Turn DialogueRetrieval-augmented generation → Conversational and speech RAG
Improving LLM-powered Recommendations with Personalized InformationLLM reasoning → LLM reasoning for recommendation
AgentCF++: Memory-enhanced LLM-based Agents for Popularity-aware Cross-domain RecommendationsLLM agents and tool use → Agentic recommender systems
Inner Thinking Transformer: Leveraging Dynamic Depth Scaling to Foster Adaptive Internal ThinkingLLM reasoning → Test-time compute and efficient thinking
Quantifying Memorization and Parametric Response Rates in Retrieval-Augmented Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning +1
Proving Olympiad Inequalities by Synergizing LLMs and Symbolic ReasoningLLM reasoning → Mathematical reasoning
ArtMentor: AI-Assisted Evaluation of Artworks to Explore Multimodal Large Language Models CapabilitiesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
In-Place Updates of a Graph Index for Streaming Approximate Nearest Neighbor SearchRetrieval-augmented generation → Vector search and ANN indexes
Scoring Verifiers: Evaluating Synthetic Verification for Code and ReasoningLLM reasoning → Reinforcement learning for reasoning +1
Learning to explore when mistakes are not allowedDeep reinforcement learning methods → Safe and constrained RL
From Correctness to Comprehension: AI Agents for Personalized Error Diagnosis in EducationLLM reasoning → Mathematical reasoning +1
VITAL: A New Dataset for Benchmarking Pluralistic Alignment in HealthcareLLM alignment / RLHF / preference optimisation → Medical LLM alignment
Agentic AI Software Engineers: Programming with TrustLLM agents and tool use → Software engineering coding agents
GIMMICK -- Globally Inclusive Multimodal Multitask Cultural Knowledge BenchmarkingMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
SCALAR: Scientific Citation-based Live Assessment of Long-context Academic ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning
Emergence of the Primacy Effect in Structured State-Space ModelsState-space sequence models (S4, Mamba) → Core SSM architectures for language
Direct Value Optimization: Improving Chain-of-Thought Reasoning in LLMs with Refined ValuesLLM reasoning → Mathematical reasoning
TrustRAG: An Information Assistant with Retrieval Augmented GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
Hierarchical RL-MPC for Demand Response SchedulingDeep reinforcement learning methods → RL for energy, networks and operations
Medical Image Classification with KAN-Integrated Transformers and Dilated Neighborhood AttentionState-space sequence models (S4, Mamba) → Mamba for medical image analysis
C2T: A Classifier-Based Tree Construction Method in Speculative DecodingEfficient LLM inference and serving → Speculative decoding
Qorgau: Evaluating LLM Safety in Kazakh-Russian Bilingual ContextsLLM safety and security → Safety alignment, refusal and guardrails
REFIND at SemEval-2025 Task 3: Retrieval-Augmented Factuality Hallucination Detection in Large Language ModelsRetrieval-augmented generation → Hallucination detection in RAG
Efficient Safety Retrofitting Against Jailbreaking for LLMsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Multi-Target Radar Search and Track Using Sequence-Capable Deep Reinforcement LearningDeep reinforcement learning methods → RL for robot navigation and UAV control
Unraveling the Localized Latents: Learning Stratified Manifold Structures in LLM Embedding Space with Sparse Mixture-of-ExpertsMixture-of-experts → MoE large language models
ETS: Efficient Tree Search for Inference-Time ScalingLLM reasoning → Test-time compute and efficient thinking +1
RestoreGrad: Signal Restoration Using Conditional Denoising Diffusion Models with Jointly Learned PriorDiffusion models → Inverse problems and image restoration
Model Evolution Framework with Genetic Algorithm for Multi-Task Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
Are Large Language Models In-Context Graph Learners?LLM reasoning → Knowledge- and tool-augmented reasoning +1
STaR-SQL: Self-Taught Reasoner for Text-to-SQLLLM reasoning → Knowledge- and tool-augmented reasoning
Exploiting Prefix-Tree in Structured Output Interfaces for Enhancing Jailbreak AttackingLLM safety and security → Jailbreaking
MobileViM: A Light-weight and Dimension-independent Vision Mamba for 3D Medical Image AnalysisState-space sequence models (S4, Mamba) → Mamba for medical image analysis
MILE: Model-based Intervention LearningOffline RL and imitation learning from data → Imitation learning methods
SPPD: Self-training with Process Preference Learning Using Dynamic Value MarginLLM reasoning → Reinforcement learning for reasoning +2
VLAS: Vision-Language-Action Model With Speech Instructions For Customized Robot ManipulationVision-language-action models → Generalist VLA models and benchmarks
PLDR-LLMs Learn A Generalizable Tensor Operator That Can Replace Its Own Deep Neural Net At InferenceEfficient LLM inference and serving → KV cache optimization
Collision-Aware Object-Goal Visual Navigation via Two-Stage Deep Reinforcement LearningDeep reinforcement learning methods → RL for robot navigation and UAV control
What are Models Thinking about? Understanding Large Language Model Hallucinations "Psychology" through Model Inner State AnalysisLLM reasoning → Hallucination detection and mitigation

All 363