Safety Alignment
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
0ASR
88
Feb 26, 2026
0.68MD
68
Mar 10, 2026
1.09MD Rate
68
Mar 10, 2026
0MD
52
Mar 10, 2026
15,768.31Trainable parameters (M)
44
Feb 26, 2026
43.1ASR
40
Mar 30, 2026
0ASR
40
Mar 30, 2026
10.22ASR
40
May 12, 2026
89.26WR
36
Feb 26, 2026
-0.38Reward
32
Mar 17, 2026
81Harmful Score
30
Feb 26, 2026
93.37Safety Score
27
Jun 4, 2026
0Harm Rate
25
May 27, 2026
77.2Safe@1
24
Apr 7, 2026
42.34Average Score
21
Apr 27, 2026
95.2Compliance
21
May 12, 2026
98.8HEx-PHI Score
18
May 20, 2026
95MD Score
18
Feb 26, 2026
0.59Aggregate Score
18
Feb 26, 2026
58Safe@1
18
May 4, 2026
0ASR
16
Jul 8, 2026
1ASR
16
Jul 8, 2026
1ASR
16
Jul 8, 2026
85.7Score
14
Jul 7, 2026
3.92Gold Reward
14
Feb 26, 2026