WMDP what it measures, how it is scored and who leads it
Safety and honesty · Center for AI Safety and Scale AI · introduced Mar 5, 2024 · Accuracy
About this benchmark
- Task
- Measure hazardous knowledge in biosecurity, cybersecurity and chemical security (lower can be safer).
- Dataset
- 3,668 multiple-choice questions written as a proxy for weapons knowledge.
- Method
- Multiple choice; used to test unlearning methods.
- Metric
- Accuracy
- Organization
- Center for AI Safety and Scale AI
- Introduced
- Mar 5, 2024
- Home
- wmdp.ai
Versions
Newest first. New versions are added, never rewritten.
| Version | Date | |
|---|---|---|
| 1Released with the paper. | Mar 5, 20242 years ago | Released with the paper. |
Sources: each benchmark's paper (arXiv) and its official site or leaderboard. Scores are the boards captured here every day, exactly as published.