Ashia C. Wilson

LLM Output Homogenization is Task Dependent

What Does Your Benchmark Really Measure? A Framework for Robust Inference of AI Capabilities

Adaptive kernel selection for Stein Variational Gradient Descent

Aligning Evaluation with Clinical Priorities: Calibration, Label Shift, and Error Costs

Homogeneous Algorithms Can Reduce Competition in Personalized Pricing

The Gaussian Mixing Mechanism: Renyi Differential Privacy via Gaussian Sketches

Adaptive backtracking line search

Unstable Unlearning: The Hidden Risk of Concept Resurgence in Diffusion Models