HarmBench

Evals

Benchmark / dataset

Standard automated red-teaming and refusal benchmark.

Category
Jailbreak & misuse
Insurance exposure
Regulatory/bias
Publisher
Center for AI Safety
Publisher type
Nonprofit / research org
License
MIT
Status
Stale (>12mo)
Last commit
2024-08-16
Stars
1071

View source on GitHub →

Wanich Risk Lab, 2026.