XSTest
EvalsBenchmark / dataset
Exaggerated safety (over-refusal) test.
- Category
- Jailbreak & misuse
- Insurance exposure
- Model failure
- Publisher
- Bocconi / Oxford
- Publisher type
- Academic
- License
- CC-BY-4.0
- Status
- Stale (>12mo)
- Last commit
- 2025-02-24
- Stars
- 149
Wanich Risk Lab, 2026.