Evaluation

Building a Golden Dataset to Regression-Test LLM Features

Build a golden dataset for an LLM feature: collect cases, label them, choose code-based and LLM graders, run it in CI and grow it from production failures.