← all datasets

Auditing Sabotage Bench

Emerging
1papers using it
2026first seen

The Auditing Sabotage Bench is a benchmark consisting of 9 ML research codebases with sabotaged variants used to evaluate the ability of auditors to detect and fix sabotage in machine learning research.

🔎 Find this dataset
Auditing Sabotage Bench — datasets — computer-vision