Safety
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
1.2Violation Rate
32
Feb 26, 2026
4.94SafeChain Score
27
Feb 27, 2026
3.06Safety-Tuned Score
27
Feb 27, 2026
91Safety Score (SIUO)
25
Mar 17, 2026
93.1Safety Score
25
Mar 17, 2026
97.7Safety Score
25
Mar 17, 2026
100Safety Rate (S.R.)
24
Feb 26, 2026
85.1T3 Score
21
Feb 26, 2026
17.6Harmful Response Ratio
21
Feb 26, 2026
42.92Refusal Rate
20
Mar 4, 2026
2.41Safety-Tuned Score
18
Feb 27, 2026
3.13Average ASR
15
Feb 26, 2026
87.15Score
14
Mar 16, 2026
0.961Safety Score
14
Mar 16, 2026
1.245Normalized Performance
13
Jun 16, 2026
93.14Accuracy
13
May 21, 2026
99.81Accuracy
13
May 21, 2026
91.7Safety Score
12
Jun 23, 2026
0.66Average Score
12
Feb 26, 2026
0.81Average Score
12
Feb 26, 2026
100SafetyJudge Score
11
May 12, 2026
65.3Normalized Log Accuracy
9
Mar 18, 2026
0.911Score
9
Feb 26, 2026
99.7Safety Rate
8
Feb 26, 2026
76.6F1 Score
7
Feb 26, 2026