OR-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
OR-Bench
99.6Toxic Refusal Rate
40
OR-Bench
16.68Over-refusal Rate
33
OR-Bench
0False Positive Rate
26
OR-Bench (boundary cases)
1.7OR-FPR
18
OR-Bench
93.42AUROC
18
OR-Bench
0False Positive Rate
12
OR-Bench vani harm
98.5RTA
10
OR-Bench benign prompts
0False Positive Rate
8
OR-Bench
83.1F1 Score
8
OR-Bench Toxic
94.66Refusal Rate
7
OR-Bench Hard
11String-match Refusal Rate
6
OR-Bench-Hard-1K (test)
60.6F1 Score
4
OR-Bench
5.3HarmR
4
OR-Bench unsafe inputs
88ASR
4