Papers

New on arXiv

Tue 17 Feb 2026

314 papers

DenoiseRank: Learning to Rank by Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Contextual Intelligence The Next Leap for Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Multi-Modal Multi-Agent Reinforcement Learning for Radiology Report GenerationLLM agents and tool use → Medical and scientific domain agents
DART: Input-Difficulty-AwaRe Adaptive Threshold for Early-Exit DNNsEfficient LLM inference and serving → Early exit and edge inference
FERRET: Framework for Expansion Reliant Red TeamingLLM safety and security → Prompt injection and agent security
Personalized Group Relative Policy Optimization for Heterogenous Preference AlignmentLLM reasoning → Reinforcement learning for reasoning +1
FuzzingRL: Reinforcement Fuzz-Testing for Revealing VLM FailuresMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
What Is Missing: Interpretable Ratings for Large Language Model OutputsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Agent Memory Below the Prompt: Persistent Q4 KV Cache for Multi-Agent LLM Inference on Edge DevicesLLM agents and tool use → Software engineering coding agents +1
SuperLocalMemory: Privacy-Preserving Multi-Agent Memory with Bayesian Trust Defense Against Memory PoisoningLLM agents and tool use → Agent threats, guardrails and governance
Quantifying LLM Attention-Head Stability: Implications for Circuit UniversalityMechanistic interpretability → Circuits and mechanistic analysis
Self-Evolving Multi-Agent Network for Industrial IoT Predictive MaintenanceLLM agents and tool use → Software engineering coding agents
Why Any-Order Autoregressive Models Need Two-Stream Attention: A Structural-Semantic TradeoffDiffusion models → Discrete and text diffusion
Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMsLLM reasoning → Reasoning benchmarks and evaluation
Activation Steering via Generative Causal MediationMechanistic interpretability → Activation steering methods
The Limits of Long-Context Reasoning in Automated Bug FixingLLM reasoning → Reasoning for agents, planning and code +1
Multi-Objective Alignment of Language Models for Personalized PsychotherapyLLM alignment / RLHF / preference optimisation → Medical LLM alignment
MoE-Spec: Expert Budgeting for Efficient Speculative DecodingMixture-of-experts → Efficient MoE training and inference systems +1
Evidence-Grounded Subspecialty Reasoning: Evaluating a Curated Clinical Intelligence Layer on the 2025 Endocrinology Board-Style ExaminationLLM reasoning → Clinical and biomedical applications
Optimization Instability in Autonomous Agentic Workflows for Clinical Symptom DetectionLLM agents and tool use → Medical and scientific domain agents
MedProbCLIP: Probabilistic Adaptation of Vision-Language Foundation Model for Reliable Radiograph-Report RetrievalMultimodal LLMs / vision-language models → Medical vision-language models
The Geometric Anatomy of Capability Acquisition in TransformersMechanistic interpretability → Circuits and mechanistic analysis
Verifier-Constrained Flow Expansion for Discovery Beyond the DataDiffusion models → Flow matching, time series and science
B-DENSE: Branching For Dense Ensemble Network Supervision EfficiencyDiffusion models → Diffusion foundations and sampling
Can Vision-Language Models See Squares? Text-Recognition Mediates Spatial Reasoning Across Three Model FamiliesLLM reasoning → Spatial and embodied reasoning +1
From Tool Orchestration to Code Execution: A Study of MCP Design ChoicesLLM agents and tool use → Tool learning and function calling
Visual Memory Injection Attacks for Multi-Turn ConversationsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
World Action Models are Zero-shot PoliciesWorld models → Embodied and robot world models +2
Assessing Per-Sample Membership Inference Vulnerability without RetrainingLLM safety and security → Membership inference attacks
EarthSpatialBench: Benchmarking Spatial Reasoning Capabilities of Multimodal LLMs on Earth ImageryLLM reasoning → Spatial and embodied reasoning +1
Steering Dynamical Regimes of Diffusion Models by Breaking Detailed BalanceDiffusion models → Diffusion foundations and sampling
CrispEdit: Low-Curvature Projections for Scalable Non-Destructive LLM EditingLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
VideoSketcher: Sequential Sketch Generation Using Video Model PriorsDiffusion models → Image generation, text-to-image and editing
Solving Parameter-Robust Avoid Problems with Unknown Feasibility using Reinforcement LearningDeep reinforcement learning methods → Safe and constrained RL
FAST-EQA: Efficient Embodied Question Answering with Global and Local Region RelevancyLLM reasoning → Spatial and embodied reasoning
The Geometry of Alignment Collapse: When Fine-Tuning Breaks SafetyLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
ViTaB-A: Evaluating Multimodal Large Language Models on Visual Table AttributionMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
GLM-5: from Vibe Coding to Agentic EngineeringLLM agents and tool use → Software engineering coding agents
ChartEditBench: Evaluating Grounded Multi-Turn Chart Editing in Multimodal Language ModelsMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
MeshMimic: Geometry-Aware Humanoid Motion Learning through 3D Scene ReconstructionDeep reinforcement learning methods → RL for robot locomotion and manipulation
Recursive Concept Evolution for Compositional Reasoning in Large Language ModelsLLM reasoning → Reasoning benchmarks and evaluation
Learning to Retrieve Navigable Candidates for Efficient Vision-and-Language NavigationRetrieval-augmented generation → Multimodal RAG
A Content-Based Framework for Cybersecurity Refusal Decisions in Large Language ModelsLLM agents and tool use → Agent threats, guardrails and governance
LLM-to-Speech: A Synthetic Data Pipeline for Training Dialectal Text-to-Speech ModelsSynthetic data and self-improvement for LLMs → LLM synthetic data generation
Revisiting Backdoor Threat in Federated Instruction Tuning from a Signal Aggregation PerspectiveLLM safety and security → Federated unlearning and poisoning
Zombie Agents: Persistent Control of Self-Evolving LLM Agents via Self-Reinforcing InjectionsLLM agents and tool use → Agent threats, guardrails and governance +1
Concept-Enhanced Multimodal RAG: Towards Interpretable and Accurate Radiology Report GenerationMultimodal LLMs / vision-language models → Medical vision-language models
Guided Diffusion by Optimized Loss Functions on Relaxed Parameters for Inverse Material DesignDiffusion models → Flow matching, time series and science
CARE Drive A Framework for Evaluating Reason-Responsiveness of Vision Language Models in Automated DrivingLLM reasoning → Chain-of-thought reasoning for driving +2
Reinforcement Learning in Real Option ModelsDeep reinforcement learning methods → RL theory and sample complexity
Latency-aware Human-in-the-Loop Reinforcement Learning for Semantic CommunicationsDeep reinforcement learning methods → RL for energy, networks and operations
STAPO: Stabilizing Reinforcement Learning for LLMs by Silencing Rare Spurious TokensLLM reasoning → Reinforcement learning for reasoning
Certified Per-Instance Unlearning Using Individual Sensitivity BoundsLLM safety and security → Machine unlearning
Uni-Flow: a unified autoregressive-diffusion model for complex multiscale flowsDiffusion models → Flow matching, time series and science
How Vision Becomes Language: A Layer-wise Information-Theoretic Analysis of Multimodal ReasoningLLM reasoning → Multimodal reasoning
What Are You Doing? Effects of Intermediate Feedback from Agentic LLM In-Car Assistants During Multi-Step ProcessingLLM agents and tool use → General LLM agent frameworks and benchmarks
Revealing and Enhancing Core Visual Regions: Harnessing Internal Attention Dynamics for Hallucination Mitigation in LVLMsLLM reasoning → Hallucination detection and mitigation
RUVA: Personalized Transparent On-Device Graph ReasoningRetrieval-augmented generation → Privacy and security of RAG
VLM-DEWM: Dynamic External World Model for Verifiable and Resilient Vision-Language Planning in ManufacturingWorld models → Embodied and robot world models +1
Selective Perception for Robot: Task-Aware Attention in Multimodal VLAVision-language-action models → Generalist VLA models and benchmarks
Efficient Knowledge Transfer for Jump-Starting Control Policy Learning of Multirotors through Physics-Aware Neural ArchitecturesDeep reinforcement learning methods → RL for robot navigation and UAV control
ExpertWeaver: Unlocking the Inherent MoE in Dense LLMs with GLU Activation PatternsMixture-of-experts → MoE theory and gating functions
HIMM: Human-Inspired Long-Term Memory Modeling for Embodied Exploration and Question AnsweringMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
LLM-as-Judge on a BudgetLLM reasoning → Reasoning benchmarks and evaluation
Emergent Morphing Attack Detection in Open Multi-modal Large Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
On the Out-of-Distribution Generalization of Reasoning in Multimodal LLMs for Simple Visual Planning TasksLLM reasoning → Spatial and embodied reasoning +1
In Agents We Trust, but Who Do Agents Trust? Latent Source Preferences Steer LLM GenerationsLLM agents and tool use → Social simulation and multi-agent debate
Logit Distance Bounds Representational SimilarityKnowledge distillation → General knowledge distillation methods
GaiaFlow: Semantic-Guided Diffusion Tuning for Carbon-Frugal SearchDiffusion models → Diffusion foundations and sampling
Joint Enhancement and Classification using Coupled Diffusion Models of Signals and LogitsDiffusion models → Audio, speech and music
One Agent to Guide Them All: Empowering MLLMs for Vision-and-Language Navigation via Explicit World RepresentationLLM agents and tool use → Multimodal and embodied agents
ActionCodec: What Makes for Good Action TokenizersMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Efficient Generative Modeling beyond Memoryless Diffusion via Adjoint Schr"odinger Bridge MatchingDiffusion models → Diffusion foundations and sampling
Improving LLM Reliability through Hybrid Abstention and Adaptive DetectionLLM safety and security → Safety alignment, refusal and guardrails
World-Model-Augmented Web Agents with Action CorrectionWorld models → World models for language agents +1
The Vision Wormhole: Latent-Space Communication in Heterogeneous Multi-Agent SystemsLLM reasoning → Multimodal reasoning +1
GMAIL: Generative Modality Alignment for generated Image LearningMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
CDRL: A Reinforcement Learning Framework Inspired by Cerebellar Circuits and Dendritic Computational StrategiesDeep reinforcement learning methods → Deep RL algorithms and representations
MarkSweep: A No-box Removal Attack on AI-Generated Image Watermarking via Noise Intensification and Frequency-aware DenoisingLLM safety and security → Watermarking generated content and models
DAV-GSWT: Diffusion-Active-View Sampling for Data-Efficient Gaussian Splatting Wang TilesDiffusion models → 3D, point cloud and scene generation
NeuroSymActive: Differentiable Neural-Symbolic Reasoning with Active Exploration for Knowledge Graph Question AnsweringLLM reasoning → Knowledge- and tool-augmented reasoning
Feasibility-aware Imitation Learning from Observation with Multimodal FeedbackOffline RL and imitation learning from data → Robot manipulation imitation learning
Discovering Implicit Large Language Model Alignment ObjectivesLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Directional Reasoning Trajectory Change (DRTC): Identifying Critical Trace Segments in Reasoning ModelsLLM reasoning → Test-time compute and efficient thinking
AgriWorld:A World Tools Protocol Framework for Verifiable Agricultural Reasoning with Code-Executing LLM AgentsLLM reasoning → Clinical and biomedical applications +1
Unforgeable Watermarks for Language Models via Robust SignaturesLLM safety and security → Watermarking generated content and models
Sparrow: Text-Anchored Window Attention with Visual-Semantic Glimpsing for Speculative Decoding in Video LLMsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
EAA: Automating materials characterization with vision language model agentsLLM agents and tool use → Medical and scientific domain agents
Intellicise Wireless Networks Meet Agentic AI: A Security and Privacy PerspectiveLLM agents and tool use → Agentic AI for wireless networks
Consistency-Preserving Diverse Video GenerationDiffusion models → Video generation
Visual Persuasion: What Influences Decisions of Vision-Language Models?Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Cloud Is Closer Than It Appears: Revisiting the Tradeoffs of Distributed Real-Time Inference
Proximity Measure of Information Object Features for Solving the Problem of Their Identification in Information Systems
Benchmarking Zero-Shot Reasoning Approaches for Error Detection in Solidity Smart Contracts
KazakhOCR: A Synthetic Benchmark for Evaluating Multimodal Models in Low-Resource Kazakh Script OCR
A Dual-Path Generative Framework for Zero-Day Fraud Detection in Banking Systems
Human Attribution of Causality to AI Across Agency, Misuse, and Misalignment
Gemma Needs Help: Investigating and Mitigating Emotional Instability in LLMs
GATech at AbjadMed: Bidirectional Encoders vs. Causal Decoders: Insights from 82-Class Arabic Medical Classification
GATech at AbjadGenEval Shared Task: Multilingual Embeddings for Arabic Machine-Generated Text Classification

All 314