Papers

New on arXiv

Mon 22 Sep 2025

415 papers

Jailbreaking LLMs via Semantically Relevant Nested Scenarios with Targeted Toxic KnowledgeLLM safety and security → Jailbreaking
ReGeS: Reciprocal Retrieval-Generation Synergy for Conversational Recommender SystemsRetrieval-augmented generation → RAG for recommendation
Design and Implementation of a Secure RAG-Enhanced AI Chatbot for Smart Tourism Customer Service: Defending Against Prompt Injection Attacks -- A Case Study of Hsinchu, TaiwanLLM safety and security → Membership inference attacks
Discrete-Time Diffusion-Like Models for Speech SynthesisDiffusion models → Audio, speech and music
Robotic Skill Diversification via Active Mutation of Reward Functions in Reinforcement Learning During a Liquid Pouring TaskDeep reinforcement learning methods → RL for robot locomotion and manipulation
Zero-Shot Visual Deepfake Detection: Can AI Predict and Prevent Fake Content Before It's Created?Diffusion models → Image generation, text-to-image and editing
CogniLoad: A Synthetic Natural Language Reasoning Benchmark With Tunable Length, Intrinsic Difficulty, and Distractor DensityLLM reasoning → Reasoning benchmarks and evaluation
Diffusion Policies with Offline and Inverse Reinforcement Learning for Promoting Physical Activity in Older Adults Using Wearable SensorsDiffusion models → Diffusion and flow policies for robotics +1
Latent Action Pretraining Through World ModelingWorld models → Embodied and robot world models +1
Losing the Plot: How VLM responses degrade on imperfect chartsMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Instruction-Following Evaluation in Function Calling for Large Language ModelsLLM agents and tool use → Tool learning and function calling
VoxGuard: Evaluating User and Attribute Privacy in Speech via Membership Inference AttacksLLM safety and security → Membership inference attacks
Measurement Score-Based MRI Reconstruction with Automatic Coil Sensitivity EstimationDiffusion models → Medical imaging and segmentation
Towards Provable Emergence of In-Context Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Evaluating the Safety and Skill Reasoning of Large Reasoning Models Under Compute ConstraintsLLM reasoning → Test-time compute and efficient thinking
TinyBEV: Cross Modal Knowledge Distillation for Efficient Multi Task Bird's Eye View Perception and PlanningKnowledge distillation → Detection and dense prediction distillation
Speculate Deep and Accurate: Lossless and Training-Free Acceleration for Offloaded LLMs via Substitute Speculative DecodingEfficient LLM inference and serving → Speculative decoding
CoRaCMG: Contextual Retrieval-Augmented Framework for Commit Message GenerationRetrieval-augmented generation → RAG for code
Brittleness and Promise: Knowledge Graph Based Reward Modeling for Diagnostic ReasoningLLM reasoning → Clinical and biomedical applications +1
Exploiting Tree Structure for Credit Assignment in RL Training of LLMsLLM reasoning → Reinforcement learning for reasoning
Towards General Computer Control with Hierarchical Agents and Multi-Level Action SpacesLLM agents and tool use → GUI and mobile agents
Seg4Diff: Unveiling Open-Vocabulary Segmentation in Text-to-Image Diffusion TransformersDiffusion models → Medical imaging and segmentation
UniPixel: Unified Object Referring and Segmentation for Pixel-Level Visual ReasoningLLM reasoning → Multimodal reasoning +1
OnePiece: Bringing Context Engineering and Reasoning to Industrial Cascade Ranking SystemLLM reasoning → LLM reasoning for recommendation
Optimization without Future Compromises? Decentralized Coordination via Collective and Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
Structuring The Future: Diffusion LLM Speculative Decoding via Calibrated Draft GraphsEfficient LLM inference and serving → Diffusion LLM and MoE inference
Reasoning Core: A Scalable RL Environment for LLM Symbolic ReasoningLLM reasoning → Reinforcement learning for reasoning
Improving Large Language Models Function Calling and Interpretability via Guided-Structured TemplatesLLM reasoning → Prompting and chain-of-thought reasoning +1
ARK-V1: An LLM-Agent for Knowledge Graph Question Answering Requiring Commonsense ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning +1
Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLMsLLM safety and security → Safety alignment, refusal and guardrails +1
TempSamp-R1: Effective Temporal Sampling with Reinforcement Fine-Tuning for Video LLMsMultimodal LLMs / vision-language models → Video understanding with MLLMs
V2V-GoT: Vehicle-to-Vehicle Cooperative Autonomous Driving with Multimodal Large Language Models and Graph-of-ThoughtsLLM reasoning → Chain-of-thought reasoning for driving +2
HuMam: Humanoid Motion Control via End-to-End Deep Reinforcement Learning with MambaState-space sequence models (S4, Mamba) → Video, motion and generative Mamba models +1
Hybrid Reputation Aggregation: A Robust Defense Mechanism for Adversarial Federated Learning in 5G and Edge Network EnvironmentsLLM safety and security → Federated unlearning and poisoning
Prepare Before You Act: Learning From Humans to Rearrange Initial StatesOffline RL and imitation learning from data → Robot manipulation imitation learning
NeuS-QA: Grounding Long-Form Video Understanding in Temporal Logic and Neuro-Symbolic ReasoningLLM reasoning → Multimodal reasoning +1
Beyond Diagnosis: Evaluating Multimodal LLMs for Pathology Localization in Chest RadiographsMultimodal LLMs / vision-language models → Medical vision-language models
Through the Lens of Human-Human Collaboration: A Configurable Research Platform for Exploring Human-Agent CollaborationLLM agents and tool use → General LLM agent frameworks and benchmarks
M3ET: Efficient Vision-Language Learning for Robotics based on Multimodal Mamba-Enhanced TransformerState-space sequence models (S4, Mamba) → Vision Mamba backbones
StableGuard: Towards Unified Copyright Protection and Tamper Localization in Latent Diffusion ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
VideoFrom3D: 3D Scene Video Generation via Complementary Image and Video Diffusion ModelsDiffusion models → 3D, point cloud and scene generation
The STAR-XAI Protocol: A Framework for Inducing and Verifying Agency, Reasoning, and Reliability in AI AgentsLLM reasoning → Reasoning for agents, planning and code +1
ComposableNav: Instruction-Following Navigation in Dynamic Environments via Composable DiffusionDiffusion models → Diffusion and flow policies for robotics
DriveDPO: Policy Learning via Safety DPO For End-to-End Autonomous DrivingEnd-to-end autonomous driving → Imitation-learning end-to-end planners +1
D-REX: A Benchmark for Detecting Deceptive Reasoning in Large Language ModelsLLM reasoning → Reasoning for agents, planning and code +2
Improving Zero-shot Sentence Decontextualisation with Content Selection and PlanningLLM reasoning → Reasoning benchmarks and evaluation
Orcust: Stepwise-Feedback Reinforcement Learning for GUI AgentLLM agents and tool use → GUI and mobile agents +2
Mitigating Strategy-Selection Bias in Reasoning for More Effective Test-Time ScalingLLM reasoning → Test-time compute and efficient thinking
Optimizing Inference in Transformer-Based Models: A Multi-Method BenchmarkDiffusion models → Diffusion foundations and sampling +1
Confidence-gated training for efficient early-exit neural networksEfficient LLM inference and serving → Early exit and edge inference
Sight Over Site: Perception-Aware Reinforcement Learning for Efficient Robotic InspectionDeep reinforcement learning methods → RL for robot navigation and UAV control
Improving After-sales Service: Deep Reinforcement Learning for Dynamic Time Slot Assignment with Commitments and Customer PreferencesDeep reinforcement learning methods → RL for energy, networks and operations
Expert-as-a-Service: Towards Efficient, Scalable, and Robust Large-scale MoE ServingMixture-of-experts → Efficient MoE training and inference systems
SocialTraj: Two-Stage Socially-Aware Trajectory Prediction for Autonomous Driving via Conditional Diffusion ModelDiffusion models → Diffusion and flow policies for robotics +1
Semantic and Visual Crop-Guided Diffusion Models for Heterogeneous Tissue Synthesis in HistopathologyDiffusion models → Medical imaging and segmentation
ContextFlow: Training-Free Video Object Editing via Adaptive Context EnrichmentDiffusion models → Video generation
Tac2Motion: Contact-Aware Reinforcement Learning with Tactile Feedback for Robotic Hand ManipulationDeep reinforcement learning methods → RL for robot locomotion and manipulation
Remote Sensing-Oriented World ModelWorld models → World models for language agents
Everyday Physics in Korean Contexts: A Culturally Grounded Physical Reasoning BenchmarkLLM reasoning → Reasoning benchmarks and evaluation
Conditional Diffusion Models for CT Image Synthesis from CBCT: A Systematic ReviewDiffusion models → Medical imaging and segmentation
LoT-Pass: Long-term-robust Image Watermarking for Image to Video GenerationLLM safety and security → Watermarking generated content and models
MotionTrans: Human VR Data Enable Motion-Level Learning for Robotic Manipulation PoliciesOffline RL and imitation learning from data → Robot manipulation imitation learning
VideoPro: Adaptive Program Reasoning for Long Video UnderstandingLLM reasoning → Multimodal reasoning
WISE: Weak-Supervision-Guided Step-by-Step Explanations for Multimodal LLMs in Image ClassificationLLM reasoning → Multimodal reasoning +1
RCTDistill: Cross-Modal Knowledge Distillation Framework for Radar-Camera 3D Object Detection with Temporal FusionKnowledge distillation → Detection and dense prediction distillation
DA-Mamba: Dialogue-aware selective state-space model for multimodal engagement estimationState-space sequence models (S4, Mamba) → Speech and audio SSMs
DINOv3-Diffusion Policy: Self-Supervised Large Visual Model for Visuomotor Diffusion Policy LearningDiffusion models → Diffusion and flow policies for robotics
EngiBench: A Benchmark for Evaluating Large Language Models on Engineering Problem SolvingLLM reasoning → Mathematical reasoning
Mechanistic Interpretability with SAEs: Probing Religion, Violence, and Geography in Large Language ModelsMechanistic interpretability → Probing deception, bias and beliefs
SD-VLM: Spatial Measuring and Understanding with Depth-Encoded Vision-Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
SISMA: Semantic Face Image Synthesis with MambaDiffusion models → Image generation, text-to-image and editing +1
Evict3R: Training-Free Token Eviction for Memory-Bounded Streaming Visual Geometry TransformersEfficient LLM inference and serving → Token pruning and merging for ViTs
MSCoRe: A Benchmark for Multi-Stage Collaborative Reasoning in LLM AgentsLLM reasoning → Reasoning for agents, planning and code +1
OmniInsert: Mask-Free Video Insertion of Any Reference via Diffusion Transformer ModelsDiffusion models → Video generation
Human vs. Agent in Task-Oriented ConversationsLLM agents and tool use → General LLM agent frameworks and benchmarks
Table2LaTeX-RL: High-Fidelity LaTeX Code Generation from Table Images via Reinforced Multimodal Language ModelsMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Interpreting Attention Heads for Image-to-Text Information Flow in Large Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Attention-based Mixture of Experts for Robust Speech Deepfake DetectionMixture-of-experts → Multimodal MoE models
GeCCo -- a Generalist Contact-Conditioned Policy for Loco-Manipulation Skills on Legged RobotsDeep reinforcement learning methods → RL for robot locomotion and manipulation
Asking a Language Model for Diverse ResponsesLLM reasoning → Test-time compute and efficient thinking
LIMI: Less is More for AgencyLLM agents and tool use → General LLM agent frameworks and benchmarks
Can LLMs Reason Over Non-Text Modalities in a Training-Free Manner? A Case Study with In-Context Representation LearningLLM reasoning → Reasoning benchmarks and evaluation
Disaggregated Prefill and Decoding Inference System for Large Language Model Serving on Multi-Vendor GPUsEfficient LLM inference and serving → Serving systems and request scheduling
An Unlearning Framework for Continual LearningLLM safety and security → Machine unlearning
Achilles' Heel of Mamba: Essential difficulties of the Mamba architecture demonstrated by synthetic dataState-space sequence models (S4, Mamba) → Core SSM architectures for language
MapCoder-Lite: Distilling Multi-Agent Coding into a Single Small LLMLLM agents and tool use → Software engineering coding agents
Privacy in Action: Towards Realistic Privacy Mitigation and Evaluation for LLM-Powered AgentsLLM agents and tool use → Agent threats, guardrails and governance
AttnComp: Attention-Guided Adaptive Context Compression for Retrieval-Augmented GenerationRetrieval-augmented generation → Generation and context integration +1
ChartHal: A Fine-grained Framework Evaluating Hallucination of Large Vision Language Models in Chart UnderstandingMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Stable Video-Driven PortraitsDiffusion models → Audio, speech and music
PRINCIPLES: Synthetic Strategy Memory for Proactive Dialogue AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
CARINOX: Inference-time Scaling with Category-Aware Reward-based Initial Noise Optimization and ExplorationLLM reasoning → Diffusion reasoning and test-time scaling +1
Training-Free Label Space Alignment for Universal Domain AdaptationMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
GeoPQA: Bridging the Visual Perception Gap in MLLMs for Geometric ReasoningLLM reasoning → Spatial and embodied reasoning +1
Multi-scale Temporal Prediction via Incremental Generation and Multi-agent CollaborationLLM agents and tool use → Multimodal and embodied agents
Single-Image Depth from Defocus with Coded Aperture and Diffusion Posterior SamplingDiffusion models → 3D, point cloud and scene generation
Evaluating Multimodal Large Language Models with Daily Composite Tasks in Home EnvironmentsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Methods for Multi-objective Optimization PID Controller for quadrotor UAVsDeep reinforcement learning methods → RL for robot navigation and UAV control
Vision Language Models Are Not (Yet) Spelling CorrectorsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
EpiCache: Episodic KV Cache Management for Long-Term Conversation on Resource-Constrained EnvironmentsEfficient LLM inference and serving → KV cache optimization

All 415