This repository contains the code for generating the benchmark results in the paper "How Surrogate Fidelity and Dataset Size Shape Benchmarking Conclusions A Case Study in Ads Ranking", to be submitted to AutoML Conference, 2026 - View it on GitHub