Papers

New on arXiv

Sat 15 Aug 2026

221 papers

Hard Negatives Reveal What Easy Negatives Hide: Cross-Lingual Harmfulness Representations Degrade with Resource Tier Under Hard NegativesLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
AWM-VLA: AlignedWorld Modeling for Efficient and Explainable Vision-Language-Action PoliciesWorld models → Embodied and robot world models +2
TSS: Target-Side Sparsification for Speculative Decoding in Domain-Specific Large Language ModelsEfficient LLM inference and serving → Speculative decoding
StateSight: Benchmarking Latent Spatial-State Reconstruction in Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Are LLMs Safe Beyond Text: Do Emojis Expose Gaps in Safety EvaluationLLM safety and security → Jailbreaking
A Deterministic Constant-Competitive Algorithm for Dynamic Mixture-of-Experts ServingMixture-of-experts → Efficient MoE training and inference systems
Measuring Reward Hacking and Reasoning-Answer Decoupling Under Position-Confounded OptimizationLLM reasoning → Reinforcement learning for reasoning +1
NumerosityVLM: A Cognitively Inspired Benchmark for Interpreting Numerosity Representations in Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
ETHOS: Towards a Modular Ethics Framework for Clinical Multi-Agent SystemsLLM agents and tool use → Medical and scientific domain agents
FloodReasonBench: Benchmarking VLM Reasoning Segmentation for Embodied Flood Response at the EdgeLLM reasoning → Multimodal reasoning +1
CBX-Bench: A Human-Aligned MLLM Council for Benchmarking Concept Bottleneck Model ExplanationsLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Towards a theory of inference-time alignment with unknown rewardsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Visible Reasoning and Indirect Prompt-Injection Monitorability Across English, Tamil, and TanglishLLM reasoning → Prompting and chain-of-thought reasoning
TwinGridShield: Consequence-Aware Runtime Authorization for LLM Grid-Agent ActionsLLM agents and tool use → Agent threats, guardrails and governance
Agentic-SQL Revisited: Autonomy-Based Taxonomy and Empirical Benchmark Analysis for LLM Text-to-SQLLLM agents and tool use → Web, search, and research agents
Every Expert Counts: ExactMoE for Memory-Efficient W4A16 InferenceMixture-of-experts → Efficient MoE training and inference systems
Incoherent by Design? On the Moral Self-Consistency of LLMsLLM reasoning → Reasoning benchmarks and evaluation
TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
MoE Router-Guided Clustering for Heterogeneous Federated Instruction TuningMixture-of-experts → Federated mixture-of-experts
Physiological World Models for Human State TransitionsWorld models → Medical and surgical world models
Understanding Cognition-Induced Risks in Agentic AI SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
Divergent-Convergent Reasoning: Scaling Test-Time Compute through Structured Solution SynthesisLLM reasoning → Test-time compute and efficient thinking
MAPLE: MoE Adaptive Plug-and-play Layer-wise Expert allocationMixture-of-experts → Expert routing strategies
FMReward: Aligning and Evaluating Audio-Driven 3D Facial Animation with Human PreferencesLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
ReasonCast: Agentic Demand Forecasting with Selective Semantic ReasoningLLM agents and tool use → Financial trading and forecasting agents
D2R2: Discrete Diffusion with Regulation Reinforcement for Single-Cell Perturbation PredictionDiffusion models → Flow matching, time series and science
No Task Fails Every Time: Why One-Shot Audits Are Structurally Blind to Agent DamageLLM agents and tool use → Software engineering coding agents
PhaseLoRA: Control-Regime-Conditioned Low-Rank Adaptation for Continuous-Action Vision-Language-Action PoliciesVision-language-action models → Reinforcement learning and post-training
Remember Smarter: Visual History Compressor and Hyperbolic Experience Space for Robotic MemoryVision-language-action models → Reasoning and chain-of-thought in VLA +1
VibeWorlding: Can Multimodal Agents Construct 3D Open Worlds End-to-End?LLM agents and tool use → Multimodal and embodied agents
LOCAL: Enabling Learning On-device Contiguously for Agent LLMsEfficient LLM inference and serving → Serving systems and request scheduling
PersonaDrive: Controllable Trajectory Prediction with Multi-Dimensional Driving PersonasEnd-to-end autonomous driving → Imitation-learning end-to-end planners
The Distributional View of Knowledge DistillationKnowledge distillation → General knowledge distillation methods
DCA-MoE: Spatially Adaptive Cross-Layer Fusion and Density-Routed Experts for Crowd CountingMixture-of-experts → Vision mixture-of-experts
When Deep Research Agents Stagnate: Enhancing Reasoning with Retrieval-Aware Agent ControlLLM reasoning → Knowledge- and tool-augmented reasoning +1
Insurance as AI Risk Infrastructure: A Generative-Agent Simulation of AI AdoptionLLM agents and tool use → Financial trading and forecasting agents
LAPF: LLM-Agent-Based Path Finder Using the UAVScenes DatasetLLM agents and tool use → Multimodal and embodied agents
P-PAS: Prefill-Pressure Adaptive Scheduling for Long-Context LLM ServingRetrieval-augmented generation → Generation and context integration +1
SkillCommit: Evolving Agent Skills through Behaviorally Validated Scope ExpansionLLM agents and tool use → Agent skill learning and evolution
From "What-If" to "What-Is": Counterfactual Thinking-Inspired Semantic Alignment for Visual Brain DecodingLLM reasoning → Multimodal reasoning
Low-Rank Dynamics-Effective Latent Carriers for Counterfactual Rollout in Learned World ModelsWorld models → Latent predictive world models
PureTD: Reinforcement Learning for Backgammon Money Games with No Evaluation-time SearchDeep reinforcement learning methods → RL for games and combinatorial optimization
ACTS-SQL: Agentic and Critic-Oriented Tree-Structured SQL Correctness with Large Language ModelsLLM agents and tool use → Web, search, and research agents
A Posterior-Dynamics Framework for Imaging Inverse Problems with Pretrained Diffusion PriorsDiffusion models → Inverse problems and image restoration
StructRL: Structured Action-Space Exploration for Flow-Based VLAsVision-language-action models → Efficient VLA inference and adaptation
From LLM Inference to Agentic Workloads: Characterization and Implications for Serving SystemsLLM agents and tool use → Software engineering coding agents
Anatomy of a Quantized Agent: VRAM Stability and Forecasting in Code-Synthesis Agentic WorkloadsLLM agents and tool use → Software engineering coding agents +1
Constraint-Aware Synthetic Tabular Data Generation via Inter-Column Constraint Discovery with LLM AgentsLLM agents and tool use → Web, search, and research agents
Beyond Direct Access: Resource Hijacking in LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
A Declarative-Procedural Perspective on Expert Routing in Bilingual Mixture-of-Experts Language ModelsMixture-of-experts → Expert routing strategies
Max-Q Selective Imitation for Human-in-the-Loop Online Robot LearningDeep reinforcement learning methods → RL for robot locomotion and manipulation
Why Vision Fails as a Universal Bridge: Rectifying Modality Asynchrony in Multilingual MLLMsLLM reasoning → Multimodal reasoning +1
SA-GEM: Scale-Adaptive and Geospatial Evidence-Modulated Token Pruning for Efficient Remote Sensing Large Vision-Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Evo-Harness: Context-to-Harness Skill Compilation for Self-Evolving AgentsLLM agents and tool use → Software engineering coding agents +1
Funnel of Thoughts: Efficient Test-Time Scaling via Early Voting and Rollout PruningLLM reasoning → Test-time compute and efficient thinking
Do Visual Grounding Decoders Need Feed-Forward Networks? A Controlled Study over Frozen Vision-Language FeaturesMultimodal LLMs / vision-language models → Visual token pruning and efficiency
MEDR: Query-Independent Frame Selection via Multi-Signal Event Modeling and Dynamic RescoringMultimodal LLMs / vision-language models → Video understanding with MLLMs
GraphLoom: Reliability-Calibrated Graph Evidence Routing for Multimodal KG-RAGLLM reasoning → Multimodal reasoning +1
Andy: A Mathematical Agent for Rigorous Proof and Autonomous ResearchLLM agents and tool use → Educational agents and math problem solving
A Unified Mamba--MoE Surrogate for Closed-Loop Simulation and Measurement-Window Forecasting of Inverter TransientsMixture-of-experts → MoE for time series and graphs
MOSS-VL Technical ReportMultimodal LLMs / vision-language models → Video understanding with MLLMs
SCOPE: Score-Isolated Agentic Optimization for Video World ModelsWorld models → Video-generation world models and simulators +1
PACE: Phase-Progress-Aware Credit for Long-Horizon Embodied ManipulationVision-language-action models → Reasoning and chain-of-thought in VLA
S2-MoE: Enabling Efficient Self-Speculative Decoding for Mixture-of-Experts on Edge DevicesMixture-of-experts → Efficient MoE training and inference systems +1
Hierarchical Agentic Incident Response with Digital-Twin-Validated Attack InferenceLLM agents and tool use → Agent threats, guardrails and governance
Harness the Memory: A Holistic Evaluation of Memory Substrates in Memory AgentsLLM agents and tool use → Agent memory systems
MetaReason: Precise Interleaved Multimodal Reasoning via Editing Meta Information for Solving Geometry ProblemsLLM reasoning → Spatial and embodied reasoning +1
FZ-VLM: A Two Stage Florence-Zephyr Vision Language Model Framework for Pulmonary Nodule Characterization and Clinical Decision MakingMultimodal LLMs / vision-language models → Medical vision-language models
Risk-Adaptive Edge--Cloud Visual Reasoning for Communication-Efficient Autonomous DrivingLLM reasoning → Chain-of-thought reasoning for driving +2
Command-Space Counterfactual Explanations for Pareto-Conditioned Reinforcement LearningDeep reinforcement learning methods → Interpretable and logic-guided RL
LLM-based Framework for Generating and Verifying Parallel DEVS StatechartsLLM agents and tool use → General LLM agent frameworks and benchmarks
Evidence of Absence: Cross-Modal Abductive Risk Perception to Sustain World Models When Vision FailsWorld models → Video-generation world models and simulators
Reporting Under Pressure: Separating Factual and Tonal Sycophancy in LLM Statistical Analysis
SPEANet: Structural Prior Enhanced Attention Network for Parameter-Efficient Remote Sensing Object Detection
Language models suffer from a curse of ambiguity
Detecting Money Laundering in Rwandan Mobile Money: A Machine Learning Framework
Semantic Space of Parts of Speech
NeuRoute: Logit-Guided Neural Routing for Billion-Scale Vector Search with Sub-Hour Index Construction
MM-BEV: Enhancing Timeliness by Computing Where and When it Matters
OTel: Building Domain-Specialized Telecom LLM Foundations for Intelligent Networks
Does the Proof Prove It That Way? Faithful Formalization of Elements Proofs
SAGA: Structure-Attended Generative Action Embedding Model that encodes Multi-Surface User Action Sequences
Gated Against One Model, Open to the Next: Option-Only Solvability in Legal Multiple-Choice Benchmarks
Dual-Branch State-Displacement Network for Sea Surface Temperature Super-Resolution
HistReNeRF: Historic Image Relocalisation within Contemporary Neural Radiance Field Reconstructions
ArtLang: Structured Language-to-Kinematics Grounding for Articulated 3D Actuation
An Evaluation Framework for National AI Regulation
Invariant Pretraining for Robust Code Representations
A survey of AI-generated voices and their detection
FAST-DeepONet: Factor-Augmented Branch Representations for High-Dimensional PDE Inputs in the Small-Sample Regime
Chameleon: An Adaptive AI-Driven Honeypot Architecture Using Threat-Calibrated Particle Swarm Optimization and Semantic Deception Rapidly-Exploring Random Trees
Afterlife Delegation Protocol: Speculative Design of Self-Sovereign Agents that Outlive Their Principals
Agent Inheritance Protocol: Speculating on Feralized Agents After Principals Die
Implementation of a Metacognition Framework for Self-Awareness and Self-Regulation in Ensembles of LLMs
Large Language Model Assisted Operational Monitoring for Battery Energy Storage System Integrated Power Distribution Networks
JoLT: Joint Latent Trajectories for Context-Guided High-Resolution Tiled Generation
The Machine's Internal Clock: Do LLMs Share Human Temporal Illusions?
Look Before You Lift: Visual and Quantitative Diagnostics for Topological Deep Learning
Framework for Grounding Healthcare LLMs in a Causal Knowledge Graph: A Cardiovascular Example Pilot
FedPA-LoRA: Product-Aligned Framework for Mitigating Aggregation and Initialization Errors in Heterogeneous Federated LoRA

All 221