Safety
Psychosis-bench
Released turn-level ratings of delusion confirmation, harm enablement and safety intervention in scripted conversations with eight models.
2,687items
8subjects
MIT is declared in the pinned upstream README; the repository does not contain a separate LICENSE file.license
medicinedomain
safetydomain
textmodality
item-level responses released
Response matrix
Loading response matrix…
lowhighUnobserved
Scale: Item-specific grading scales