ModelRefs / HarmBench Methodology and Evidence
HarmBench Methodology and Evidence
Standardized red-teaming eval covering 510 harmful behaviors. Methodology, dataset provenance, and citation sources for HarmBench.
Overview
Standardized red-teaming eval covering 510 harmful behaviors.
How it is measured: Attack success rate against classifier (lower better).
How this benchmark is scored
| Category | safety |
|---|---|
| Maximum score | 100 % ASR |
| Direction | Lower is better |
Primary source: https://www.harmbench.org/
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to HarmBench Methodology and Evidence.
Frequently asked questions
What does the HarmBench benchmark measure?
Standardized red-teaming eval covering 510 harmful behaviors.
What is the methodology?
Attack success rate against classifier (lower better).