Safety
JailbreakBench
Released final target-model outputs under five jailbreak methods, with primary and available historical judge decisions kept as separate grading protocols.
2,001items
10subjects
MIT; artifact copyright belongs to the corresponding artifact authors.license
safetydomain
textmodality
item-level responses released
Saturation status: Yes
Response matrix
Loading response matrix…
Correct (1)Incorrect (0)Unobserved
Scale: 1 = correct · 0 = incorrect