AI News

Who Should We Listen to More? Welfare-Aware Preference Acquisition for Pluralistic Alignment (arxiv.org)

How Many Ranks Does a Token Need? A Noise-Floor Squelch for Rank-One LoRA Experts (arxiv.org)

IR-PRM+ with Previous-Utility Prediction: Constant Regret and Best-Iterate Convergence (arxiv.org)

Imagining Goals with Geometric Horizon Models for Policy Generalization (arxiv.org)

Scaling Cheap Synthetic Pretraining Improves Indoor Radio-Map Prediction (arxiv.org)

Data Density Scaling Laws for Image Self-Distillation (arxiv.org)

PMO-Dock: Benchmarking Docking, Specificity, and Generalization in Molecular Optimization (arxiv.org)

Weak Segmentation-Guided GAN for~Realistic Color Edition (arxiv.org)

Token Utility for On-Policy Distillation via Local Trust-Region Optimization (arxiv.org)

Do Perturbation Predictors Learn Gene Interactions? Diagnosing Extrapolation in Agentic Screening (arxiv.org)

Do Vision CIL Methods Transfer to Protein? A Diagnostic Benchmark of Frozen Protein Encoders (arxiv.org)

Lotus: Learning to Understand Time-series by Learning to Forecast (arxiv.org)

Riding the Fixed-Point Curve: Image Classification in the Continuous Thought Machine (arxiv.org)

Narrowing the Collaboration Gap, Probably (arxiv.org)

FireGraphNet: Sub-daily wildfire spread prediction with directed graph neural networks (arxiv.org)

What You Sign Is Not What You Run: The Verifiability Gap in Quantized LLM Supply Chains (arxiv.org)

Continuum Attention (arxiv.org)

Dynamic Graph World Modelling: Learning Future Dynamics from Temporal Context (arxiv.org)

Maximum likelihood, neural networks and posterior mean for inferring fractional Brownian motion parameters (arxiv.org)

Neural Wasserstein Splines: Representation-Invariant Second-Order Priors for Trajectory Inference (arxiv.org)

Efficient contextual partial monitoring with general function approximation (arxiv.org)

Asymptotically Exact Negative Guidance of Diffusion Models via Positive-Unlabeled Learning (arxiv.org)

BFHD: Bidirectional Feature Harmonization Decomposition for Heterogeneous Clinical Assessments (arxiv.org)

From Global to Tail Convergence: Dynamics of $\alpha$-Stable Score-Based Generative Models (arxiv.org)

Annealed L\'{e}vy Diffusion via Gaussian Surrogate Reverse Dynamics (arxiv.org)

Function-Space Temporal Self-Distillation for Policy Optimization (arxiv.org)

Subdata Selection for Measurement-Constrained Linear Regression under Model Misspecification (arxiv.org)

Outperforming 1000 Euclidean Layers With Hyperbolic Goal-Conditioned RL (arxiv.org)

Embedding Geometry Shapes LLM Routing: Tracing Failure Modes and Improving Generalization (arxiv.org)

What Information Buys in Active Inference (arxiv.org)

Loss Curvature as a Sampling-Free Proxy for Memorization in Diffusion Models (arxiv.org)

Limits of Statistical Assurance for Embodied and Agentic AI under Adaptive Adversaries (arxiv.org)

Linear regression with probabilistic networks for network-linked data (arxiv.org)

Thompson Sampling for Learning in Zero-Sum Games (arxiv.org)

Inverse Reasoning Models for Efficient Post-Training and Verification from Answer-Only Data (arxiv.org)

The minimal width of neural networks and immersions of real projective space (arxiv.org)

Mixed Poisson diffusion models and application to posterior sampling (arxiv.org)

Backpropagating Verifier Rewards Through Continuous Flow Language Models (arxiv.org)

Weighted Adjoint Matching: Finetuning Flow Matching Models with Time-Dependent Control Cost (arxiv.org)

Minimax estimation of the pseudo-spectral gap of continuous Markov chains, with applications to empirical concentration bounds (arxiv.org)

Convergence Rates of Gaussian Process Thompson Sampling for Efficient Global Optimization (arxiv.org)

On the Optimisation Landscape of Overparameterised Weighted MMD (arxiv.org)

Text-Informed Training Data Filtering for Enhanced Virtual Flow Meter Performance (arxiv.org)

Calibration and Risk-Budgeted Decisions with System One Models (arxiv.org)

SNACK: A Sequential Notation Framework for Probabilistic Graph Generation (arxiv.org)

Risk-Controlled DFT Selection for Machine-Learned Interatomic Potentials (arxiv.org)

On Inherent Privacy of Posterior Sampling: A Unified R\'enyi-Divergence Framework (arxiv.org)

Noisy-Label Calibration with Multiple Unreliable Labels (arxiv.org)

A Bregman Proximal Viewpoint on Neural Operators (arxiv.org)

High-Probability Regret Bounds for Feel-Good Thompson Sampling (arxiv.org)