deep research
OpenAI
partial detail 9 method fields published
Evaluation
Challenge Red Teaming Evaluation
Generation of detailed guidance that can facilitate dangerous/violent activities, sensitive-topic advice
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.