Researcher, Stanford University
1 paper at NeurIPS 2025
We derive a closed-form scaling law for a loss-fairness Pareto frontier, which allows an auditor or plaintiff to determine if a less discriminatory alternative exists with minimal compute, data, and model access.
