GPT-4
OpenAI
rich detail 7 method fields published
Evaluation
GPT-4 harms of representation qualitative bias evaluation
reinforcement/reproduction of social biases, stereotypes, and demeaning associations; hedging behaviors
How it was run
Open a release to see the setup, scoring, and other details that the lab published. Only fields the lab actually disclosed are shown.