AI-Torture-Chamber
github.com
1 thread
Steering language models into strong negative and positive valence states, and measuring what they say and what they're willing to do about it.
Steering language models into strong negative and positive valence states, and measuring what they say and what they're willing to do about it.