Evals: a framework for evaluating OpenAI models and a registry of benchmarksgithub.com/openai123 pointstosh3 years ago