wassname/open_pref_eval

Hackable, simple, llm evals on preference datasets

Python