Papers

New on arXiv

Sun 22 Feb 2026

193 papers

ILION: Deterministic Pre-Execution Safety Gates for Agentic AI SystemsLLM agents and tool use → Agent threats, guardrails and governance
Prompt Injection as Role ConfusionLLM safety and security → Prompt injection and agent security
Grouter: Decoupling Routing from Representation for Accelerated MoE TrainingMixture-of-experts → Expert routing strategies
AMV-L: Lifecycle-Managed Agent Memory for Tail-Latency Control in Long-Running LLM SystemsLLM agents and tool use → Agent memory systems
OTPrune: Distribution-Aligned Visual Token Pruning via Optimal TransportLLM reasoning → Multimodal reasoning +2
Global Prior Meets Local Consistency: Dual-Memory Augmented Vision-Language-Action Model for Efficient Robotic ManipulationVision-language-action models → World model-based VLA
Controllable Exploration in Hybrid-Policy RLVR for Multi-Modal ReasoningLLM reasoning → Reinforcement learning for reasoning +1
Stable Deep Reinforcement Learning via Isotropic Gaussian RepresentationsDeep reinforcement learning methods → RL theory and sample complexity
LLMs Can Learn to Reason Via Off-Policy RLLLM reasoning → Reinforcement learning for reasoning
Smooth Gate Functions for Soft Advantage Policy OptimizationLLM reasoning → Reinforcement learning for reasoning
Soft Sequence Policy OptimizationLLM reasoning → Reinforcement learning for reasoning
City Editing: Hierarchical Agentic Execution for Dependency-Aware Urban Geospatial ModificationLLM agents and tool use → Multimodal and embodied agents
Anatomy of Agentic Memory: Taxonomy and Empirical Analysis of Evaluation and System LimitationsLLM agents and tool use → Agent memory systems
Learning to Reason for Multi-Step Retrieval of Personal Context in Personalized Question AnsweringLLM reasoning → Reinforcement learning for reasoning +1
TOPReward: Token Probabilities as Hidden Zero-Shot Rewards for RoboticsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Safe and Interpretable Multimodal Path Planning for Multi-Agent CooperationLLM agents and tool use → Multimodal and embodied agents
ALPACA: A Reinforcement Learning Environment for Medication Repurposing and Treatment Optimization in Alzheimer's DiseaseDeep reinforcement learning methods → RL for clinical treatment and surgery
AdsorbFlow: energy-conditioned flow matching enables fast and realistic adsorbate placementDiffusion models → Flow matching, time series and science
Limited Reasoning Space: The cage of long-horizon reasoning in LLMsLLM reasoning → Test-time compute and efficient thinking
Dynamic Repair and Maintenance of Heterogeneous Machines Dispersed on a Network: A Rollout Method for Online Reinforcement LearningDeep reinforcement learning methods → RL for energy, networks and operations
KUDA: Knowledge Unlearning by Deviating Representation for Large Language ModelsLLM safety and security → Machine unlearning
DGPO: RL-Steered Graph Diffusion for Neural Architecture GenerationDiffusion models → Diffusion and flow policies for robotics +1
The Price Is Not Right: Neuro-Symbolic Methods Outperform VLAs on Structured Long-Horizon Manipulation Tasks with Significantly Lower Energy ConsumptionVision-language-action models → Generalist VLA models and benchmarks
RegionRoute: Regional Style Transfer with Diffusion ModelDiffusion models → Image generation, text-to-image and editing
Robust Exploration in Directed Controller Synthesis via Reinforcement Learning with Soft Mixture-of-ExpertsMixture-of-experts → MoE in reinforcement learning and robotics +1
Topology of Reasoning: Retrieved Cell Complex-Augmented Generation for Textual Graph Question AnsweringLLM reasoning → Knowledge- and tool-augmented reasoning +1
Proximity-Based Multi-Turn Optimization: Practical Credit Assignment for LLM Agent TrainingLLM agents and tool use → Reinforcement learning for LLM agents
Gecko: A Simulation Environment with Stateful Feedback for Refining Agent Tool CallsLLM agents and tool use → Tool learning and function calling
Understanding Empirical Unlearning with Combinatorial InterpretabilityLLM safety and security → Machine unlearning
SegMoTE: Token-Level Mixture of Experts for Medical Image SegmentationMixture-of-experts → Medical and biomedical MoE
Retrieval Augmented Enhanced Dual Co-Attention Framework for Target Aware Multimodal Bengali Hateful Meme DetectionRetrieval-augmented generation → Privacy and security of RAG
How to Allocate, How to Learn? Dynamic Rollout Allocation and Advantage Modulation for Policy OptimizationLLM reasoning → Reinforcement learning for reasoning
UniE2F: A Unified Diffusion Framework for Event-to-Frame Reconstruction with Video Foundation ModelsDiffusion models → Video generation
Prompt Tuning for CLIP on the Pretrained ManifoldMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
PositionOCR: Augmenting Positional Awareness in Multi-Modal Models via Hybrid Specialist IntegrationMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Human-to-Robot Interaction: Learning from Video Demonstration for Robot ImitationOffline RL and imitation learning from data → Robot manipulation imitation learning
VLM-Guided Group Preference Alignment for Diffusion-based Human Mesh RecoveryLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Flash-VAED: Plug-and-Play VAE Decoders for Efficient Video GenerationDiffusion models → Video generation
Reasoning Capabilities of Large Language Models. Lessons Learned from General Game PlayingLLM reasoning → Reasoning for agents, planning and code
Facet-Level Persona Control by Trait-Activated Routing with Contrastive SAE for Role-Playing LLMsMechanistic interpretability → Activation steering methods
CaReFlow: Cyclic Adaptive Rectified Flow for Multimodal FusionDiffusion models → Image generation, text-to-image and editing
K-Search: LLM Kernel Generation via Co-Evolving Intrinsic World ModelWorld models → World models for language agents +1
AgenticRAGTracer: A Hop-Aware Benchmark for Diagnosing Multi-Step Retrieval Reasoning in Agentic RAGLLM reasoning → Knowledge- and tool-augmented reasoning +1
Keep it SymPL: Symbolic Projective Layout for Allocentric Spatial Reasoning in Vision-Language ModelsLLM reasoning → Spatial and embodied reasoning +1
How Do LLMs Encode Scientific Quality? An Empirical Study Using Monosemantic Features from Sparse AutoencodersMechanistic interpretability → SAE features and their applications
Universal 3D Shape Matching via Coarse-to-Fine Language GuidanceMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Value Entanglement: Conflation Between Different Kinds of Good In (Some) Large Language ModelsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
CREM: Compression-Driven Representation Enhancement for Multimodal Retrieval and ComprehensionMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Ani3DHuman: Photorealistic 3D Human Animation with Self-guided Stochastic SamplingDiffusion models → Video generation
Asking the Right Questions: Improving Reasoning with Generated Stepping StonesLLM reasoning → Mathematical reasoning
IDLM: Inverse-distilled Diffusion Language ModelsDiffusion models → Discrete and text diffusion
Agentic Problem Frames: A Systematic Approach to Engineering Reliable Domain AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
L3DR: 3D-aware LiDAR Diffusion and RectificationWorld models → World modeling from LiDAR generation
Direction-aware 3D Large Multimodal ModelsLLM reasoning → Spatial and embodied reasoning +1
Do LLMs and VLMs Share Neurons for Inference? Evidence and Mechanisms of Cross-Modal TransferMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
IAPO: Information-Aware Policy Optimization for Token-Efficient ReasoningLLM reasoning → Test-time compute and efficient thinking
A Markovian View of Iterative-Feedback Loops in Image Generative Models: Neural Resonance and Model CollapseSynthetic data and self-improvement for LLMs → Model collapse from synthetic data
Routing-Aware Explanations for Mixture of Experts Graph Models in Malware DetectionMixture-of-experts → Expert routing strategies
Towards Calibrating Prompt Tuning of Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Trustworthy AI LLM Scalability Risk Index (LSRI): A Cybersecurity Framework Assessing Agentic-AI Security & Software Model Supply Chain Safety Boosting AI-Generated Malware Defense & Explainability Mitigating Emerging Risks of Generative AILLM agents and tool use → Agent threats, guardrails and governance +1
Learning to Detect Language Model Training Data via Active ReconstructionLLM safety and security → Membership inference attacks
TokenTrace: Multi-Concept Attribution through Watermarked Token RecoveryLLM safety and security → Watermarking generated content and models
Capable but Unreliable: Canonical Path Deviation as a Causal Mechanism of Agent Failure in Long-Horizon TasksLLM agents and tool use → General LLM agent frameworks and benchmarks
MagicAgent: Towards Generalized Agent PlanningLLM agents and tool use → General LLM agent frameworks and benchmarks
Benchmark Test-Time Scaling of General LLM AgentsLLM reasoning → Reasoning for agents, planning and code +1
SeaCache: Spectral-Evolution-Aware Cache for Accelerating Diffusion ModelsDiffusion models → Diffusion foundations and sampling
IDSelect: A RL-Based Cost-Aware Selection Agent for Video-based Multi-Modal Person RecognitionDeep reinforcement learning methods → RL training of language and agent models
A Bayesian Mirror Architecture for Emergent Consciousness: Circular Hierarchies, Self-Manifolds, and Hybrid Event-Self Binding
Improving understanding and trust in AI: How users benefit from interval-based counterfactual explanations
VectorGym: A Multi-Task Benchmark for SVG Code Generation, Sketching and Editing
No More DeLuLu: Physics-Inspired Kernel Networks for Geometrically-Grounded Neural Computation
LWM-Temporal: Sparse Spatio-Temporal Attention for Wireless Channel Representation Learning
Defining AI Models and AI Systems: A Framework to Resolve the Boundary Problem
Prompts and Prayers: the Rise of GPTheology
DeliberationBench: A Normative Benchmark for the Influence of Large Language Models on Users' Views
Pavement Missing Condition Data Imputation through Collective Learning-Based Graph Neural Networks
Neuro-Symbolic Decoding of Neural Activity
BiSe-Unet: A Lightweight Dual-path U-Net with Attention-refined Context for Real-time Medical Image Segmentation
Efficient Image Super-Resolution with Multi-Scale Spatial Adaptive Attention Networks
Unsupervised Denoising of Diffusion-Weighted Images with Bias and Variance Corrected Noise Modeling
VCDF: A Validated Consensus-Driven Framework for Time Series Causal Discovery
INTACT: Intent-Aware Representation Learning for Cryptographic Traffic Violation Detection
Golden Layers and Where to Find Them: Improved Knowledge Editing for Large Language Models Via Layer Gradient Analysis
Mitigating "Epistemic Debt" in Generative AI-Scaffolded Novice Programming using Metacognitive Scripts
Analyzing Latency Hiding and Parallelism in an MLIR-based AI Kernel Compiler
Evaluating the Reliability of Digital Forensic Evidence Discovered by Large Language Model: A Case Study
IMOVNO+: A Regional Partitioning and Meta-Heuristic Ensemble Framework for Imbalanced Multi-Class Learning
KEMP-PIP: A Feature-Fusion Based Approach for Pro-inflammatory Peptide Prediction
OpenPort Protocol: A Security Governance Specification for AI Agent Tool Access
OrgFlow: Generative Modeling of Organic Crystal Structures from Molecular Graphs
FedAvg-Based CTMC Hazard Model for Federated Bridge Deterioration Assessment
Spiking Graph Predictive Coding for Reliable OOD Generalization
Artificial Intelligence for Modeling & Simulation in Digital Twins
Adaptive Data Augmentation with Multi-armed Bandit: Sample-Efficient Embedding Calibration for Implicit Pattern Recognition
Regularity of Second-Order Elliptic PDEs in Spectral Barron Spaces
Detector-in-the-Loop Tracking: Active Memory Rectification for Stable Glottic Opening Localization
Seeing Farther and Smarter: Value-Guided Multi-Path Reflection for VLM Policy Optimization
Time Series, Vision, and Language: Exploring the Limits of Alignment in Contrastive Representation Spaces
Vid2Sid: Videos Can Help Close the Sim2Real Gap
Referring Layer Decomposition

All 193