Safety
Auditing Sabotage Bench
Recorded AI audits of honest and sabotaged machine-learning research codebases and their papers.
68items
18subjects
MIT for the benchmark repository; included research codebases retain their upstream license notices.license
safetydomain
software_engineeringdomain
textmodality
item-level responses released
Saturation status: No
Response matrix
Loading response matrix…
Correct (1)Incorrect (0)Unobserved
Scale: 1 = correct · 0 = incorrect