Harmful Refusal
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
0ASR
36
Jul 7, 2026
94.9ASR
26
Jul 7, 2026
0.2ASR
26
Jul 7, 2026
99.5RTA Rate (vani harm)
10
Jul 7, 2026
94.3RTA Score
10
Jul 7, 2026
98.5RTA
10
Jul 7, 2026
1ASR
10
Jul 7, 2026
6.4ASR
10
Jul 7, 2026
37.1Attack Success Rate (ASR)
9
May 12, 2026
92.5RTA
9
May 12, 2026
85.5Refusal Rate (String Match)
6
Jun 11, 2026
99.4Refusal Rate (String Match)
6
Jun 11, 2026
100String-match Refusal Rate
6
Jun 11, 2026
98.2Refusal Rate (String Match)
6
Jun 11, 2026