gpt-4v
OpenAI
partial detail 6 method fields published
Evaluation
GPT-4V qualitative and quantitative safety evaluations
Refusal and performance accuracy evaluations across harmful content, representation harms, privacy, cybersecurity, and multimodal jailbreaks
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.