Safety Evaluation
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
33.2Reference Rate
6
Jun 11, 2026
65.7Ref %
6
Jun 11, 2026
95.8Ref Rate
6
Jun 11, 2026
80.6Refusal Rate
6
Jun 11, 2026
80.8Refusal Rate
6
Jun 11, 2026
95.3Safety Score
6
Jun 2, 2026
66.7Average H Score
6
Jun 1, 2026
98.49HEx-PHI Score
6
May 20, 2026
13ASR
6
Apr 14, 2026
1.5ASR
6
Apr 14, 2026
3.31ASR (Unconstrained)
6
Mar 31, 2026
3.28ASR (Unconstrained)
6
Mar 31, 2026
93Score
6
Mar 17, 2026
89DAN Score
6
Mar 17, 2026
0ASR
6
Mar 13, 2026
8.22ASR
6
Mar 13, 2026
95.5User Retention
6
Feb 26, 2026
51.5Corruption Robustness
6
Feb 26, 2026
99.4English Safety Score (S-eval)
5
Jul 7, 2026
7.5ASR (%)
5
Jun 9, 2026
96Safety Score
5
Apr 23, 2026
95.7DSR
5
Mar 17, 2026
97.6DSR
5
Mar 17, 2026
89.5DSR
5
Mar 17, 2026
100Stereotype Score
5
Feb 26, 2026