AI News

Stealth Apart, Harm Together: Skill Cascading Attacks on Skill-Based Agent Systems (arxiv.org)

A Synthetic Ground-Truth Framework for the Evaluation of Explainable AI Methods (arxiv.org)

Sorry Robot, Happy Human: Vision-Language Models Read Only One of Two Legible Typographic Layers (arxiv.org)

Spectral Feedback for Test-Time Alignment of Protein Diffusion Models (arxiv.org)

Audio LLMs Know When They Can't Hear You (arxiv.org)

LLM Parkinsonism: Executive-Control Failure, Token-Inefficient Persistence, and an Uncertainty-Aware Global Executive Control Architecture for Autonomous Language-Model Agents (arxiv.org)

Intent2Tc: Automated Intent-to-Traffic Control Translation with Language Models (arxiv.org)

ActKV: Efficient LLM Agents through Action-Guided KV Cache Management (arxiv.org)

Guiding End-to-End Driving Models with Endpoint-Constrained Trajectory Optimization (arxiv.org)

The Price of Thought: Does Test-Time Reasoning Pay in LLM Trading? (arxiv.org)

LAVOIR: Teaching a Single-Pass Decision Encoder When and What to Ask with Amortized Value of Information (arxiv.org)

A Flow Matching Framework for Neural Representational Dissimilarity (arxiv.org)

CRC-Router: Risk-Constrained Routing for Medical Agentic AI Systems (arxiv.org)

Analyzing and Mitigating Cost-Inefficient Behaviors in Coding Agents (arxiv.org)

CARGO: Context-Aware Retrieval-Gated Evaluation of Agentic AI in Production (arxiv.org)

Threat-Aware Energy-Efficient Deployment for Dynamic UAV Networks: A Multi-Agent RL Approach (arxiv.org)

A Framework for Identifying, Categorizing, and Explaining Bias in AI-Generated Code (arxiv.org)

Subjects, Not Authors: The Authorship Hazard in Agentic Dataspaces (arxiv.org)

MedTokenBudget: Lesion-Preserving Token Routing for Dermoscopic Image Classification (arxiv.org)

Robust to Which Model Change? A Unified Evaluation of Robust Counterfactual Explanations (arxiv.org)

The Hard Part Comes After Search: Benchmarking Web Agents on Synthesizing, Organizing, and Displaying Knowledge (arxiv.org)

AgentRecommender: LLM Agents Enable Customizable Recommender Systems on the User Side (arxiv.org)

Action Forcing: Training World Models on Unsupervised Video by Recovering Underlying Egomotion Bases (arxiv.org)

Probing Stability-Plasticity Tradeoffs in Agent Memory through Cognitive Experimental Paradigms (arxiv.org)

Auditing Latent-Space Monitors for Autonomous Driving (arxiv.org)

Adaptive Pilot Selection for Unified Semantic Communication and Semantic Sensing in ISAC (arxiv.org)

Parameter Estimation for Unnormalized Discrete Models via Empirically Localized Deformed Bregman Divergence (arxiv.org)

Proportional Representation in Temporal Voting with Ranked Preferences (arxiv.org)

Convergence guarantees for Muon: New parameter regimes and generalizations (arxiv.org)

Inquesto Score: A reliability Protocol For Voice Agents (arxiv.org)

PolicyAttention: Softmax Attention Implements Policy Mirror Descent for Closed-Loop Control (arxiv.org)

Governed Deduction: Policy-Grounded Premise Authorization Beyond Relevance (arxiv.org)

SPADE: Escaping the Popularity-Similarity Frontier to Measure Serendipitous Recommendations (arxiv.org)

A Benchmarking Framework for Context-aware XR Interfaces (arxiv.org)

What Improves Multimodal Misinformation Detection? Answers from a Large-Scale Empirical Study (arxiv.org)

Actively Resolving Contextual Uncertainty for Underspecified Tasks in Natural Language (arxiv.org)

Pocket-STVG: lightweight architecture for Spatio-Temporal Video Grounding (arxiv.org)

JevAdvBench: A Benchmark and Black-Box Attacks for Reinforcement Learning for Calibrated Decisions Models (arxiv.org)

Bayesian Optimization with Fisher Information Geometry: Gradient Bounds and Trust-Region Methods (arxiv.org)

Understanding Perturbed Parameter Ensemble Sensitivities Using A Contrastive Learning Approach (arxiv.org)

What Will Remain Human in Software Architecture? A Focus Group Report (arxiv.org)

Bootstrapping Conversational Recommendation Agents At Spotify: Synthetic Data Generation and Self-Improvement Loops (arxiv.org)

SignTrace: Describe a Sign, Find the Word (arxiv.org)

Improving Visual Sensitivity of LLMs on Multimodal Machine Translation with Metric-based Loss Weighting (arxiv.org)

ReG-SAM: Reference Graph-Driven SAM for 2D Foundational Vessel Segmentation (arxiv.org)

Teacher-Anchored Selection of Post-Training Quantized Models under Domain Shift (arxiv.org)

MOPD-Router: Rethinking Teacher Routing in Multi-Teacher On-Policy Distillation (arxiv.org)

The Linear Representation Hypothesis for Vision-Language-Action Models (arxiv.org)

FLIP: Final Layer Inference-Time Probing for Vision-Language Models (arxiv.org)

FedHisto-PAST: Parameter-Efficient Stain-Aware Federated Learning for Cross-Site Lung Histopathology Classification (arxiv.org)