The README currently says that: It currently wires up the 2025 evaluation challenges and defaults to grok_config_prod with limit=1 so you can smoke-test the pipeline quickly. However, the run.py is currently set up with GPT 5 Pro, so this needs to be fixed for consistency.
Activity