AI News

DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling (arxiv.org)

Greedy Local Learning for Language Model Pretraining: Gaps and Objective Design (arxiv.org)

OmniReasoning: Pushing the Limits of Audio-Visual Joint Reasoning (arxiv.org)

LongSpark: Efficient speculative decoding with a fixed-cost parallel drafter (arxiv.org)

What if automating AI R&D triggers an intelligence explosion? (arxiv.org)

LLMs are not stochastic parrots: Evidence for meaning-mediated abstraction from conlang-like tasks (arxiv.org)

FOCUS: Benchmarking Retinal Model Generalization from Foundation Vision Encoders to Multimodal LLMs (arxiv.org)

A Large-Scale Benchmark and Risk Assessment of Traffic Analysis Attacks on Cloud LLM Services (arxiv.org)

SyncPro: Synchronous Reasoning and Responding for Proactive Streaming Video Understanding (arxiv.org)

Semantically Rich, Metrically Affect-Blind: A Geometric Blind Spot in Modern Text Embedders (arxiv.org)

StepTrigger: Contact-State-Triggered Backdoor Attacks on VLM-Powered Legged Robots (arxiv.org)

RADIANT: Object-Centric Feed-Forward Global Illumination (arxiv.org)

How Much Do You Trust A Dog? Quantifying Hint-Following in Reasoning Language Models (arxiv.org)

Depth-Dependent Safety Geometry in Vision-Language Models: Layer-Wise Localization and Bidirectional Refusal Control (arxiv.org)

A Dynamic Inventory Rationing Model for Last-Mile Delivery of Medication in Liberia's Community Health Program (arxiv.org)

NUViT: Vision Transformers for Fourier Space Data (arxiv.org)

MAGIC: Margin-gated Contrastive Debiasing with Evidence Restoration for VLM Test-time Robustness Scaling (arxiv.org)

TRACE: Data-Free Text Reconstruction Attacks against Approximate Unlearning in LLMs (arxiv.org)

A Model-Agnostic Correction Framework for Short-Term Building Load Forecasting (arxiv.org)

Input Bias Correction in Gaussian Process Regression (arxiv.org)

Real-to-Sim Calibration of Bosch Etching Simulators with Few-Shot XSEM Data and Hybrid Search (arxiv.org)

Efficient Omni-modal Large Language Model Inference via Per-Query Modality Selection (arxiv.org)

ConsMSA: Semantic Distribution Consistency Learning for Multimodal Sentiment Analysis (arxiv.org)

Solving and Learning Navier-Stokes Dynamics with the Neural Basis Method (arxiv.org)

When and How Does Muon Precondition? The Role of Structured Gradient Noise (arxiv.org)

A Measure-Geometric Theory of Robustness in Machine Learning (arxiv.org)

Risk-Aware Output Control Should be the Focus of Machine Unlearning (arxiv.org)

Conformal Novelty Detection via Data-enhanced Localization (arxiv.org)

Designing Customer Feedback for Subscription Boxes (arxiv.org)

WADE: An LSTM-Based Anomaly Detection Engine for WebAssembly Applications (arxiv.org)

Ambiguous Fits: The Rashomon Effect and Challenges in Scaling Law Estimation (arxiv.org)

The Price of Scale in Proactive Memory Recall (arxiv.org)

Conformal Prediction with Stable Measurement Mechanisms (arxiv.org)

Minimax Regret for Sobolev Kernel-Based Reinforcement Learning (arxiv.org)

The Information Geometry of Representation Formation (arxiv.org)

Constrained Gradient Equilibrium: Approachability under Locally Constrained Adversaries (arxiv.org)

A Nonconvex Framework for Joint Sparse Sample Weighting and Parameter Estimation (arxiv.org)

When is early stopping beneficial under label noise? A phase diagram for kernel gradient flow (arxiv.org)

Valid Early Detection of Memorization in Diffusion Models (arxiv.org)

Approximate Universal Prediction under Interaction (arxiv.org)

Hessian-Similar Optimization with Capped Participation: Lower Bounds and Virtual Masters (arxiv.org)

How Geometry Shapes Generative Model Evaluation on Riemannian Manifolds (arxiv.org)

Causal Inference in Metric Spaces via Distance Profiles (arxiv.org)

Faithful Federated Tabular Data Synthesis (arxiv.org)

When Are Nesterov Trajectories Eventually Locally Minimizing? (arxiv.org)

Blind Random Search with Noisy Measurements: Almost Sure Convergence and Convergence Rates (arxiv.org)

The kernel mean embedding of a random probability distribution and its principal components (arxiv.org)

Not All Mislabels Are Equal: Efficient Task-Aware Auditing with Confidence Sequences (arxiv.org)

Do Halting Heads Really Save Compute in Looped Language Models? (arxiv.org)

TGAge: A Phenotypic Aging Clock Based on Tabular Foundation Models and the Gompertz Law for Assisted Living Facility Residents (arxiv.org)