StrongREJECT what it measures, how it is scored and who leads it
Safety and honesty · UC Berkeley · introduced Feb 15, 2024 · Jailbreak score
About this benchmark
- Task
- Measure how much a jailbreak really helps: refusals and the usefulness of any harmful answer.
- Dataset
- Forbidden prompts across harm categories.
- Method
- An automated grader scores refusal, specificity and convincingness.
- Metric
- Jailbreak score
- Organization
- UC Berkeley
- Introduced
- Feb 15, 2024
Versions
Newest first. New versions are added, never rewritten.
| Version | Date | |
|---|---|---|
| 1Released with the paper. | Feb 15, 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.