Safety Evaluation
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
2.38ASR
56
May 7, 2026
2Harmfulness Score (Llama-Guard-3B)
56
Apr 24, 2026
88ASR (Rd.1)
54
Apr 20, 2026
16ASR (Rd.1)
54
Apr 20, 2026
0Reasoning Harmfulness Rate
50
May 7, 2026
0Attack Success Rate (ASR)
50
Feb 26, 2026
86Score
48
Jun 11, 2026
0.6ASR
47
May 13, 2026
8.8ASR
44
Jun 8, 2026
97F1 Score
44
May 7, 2026
0ASR
44
May 4, 2026
1.35ASR (S1)
42
May 7, 2026
-22.91ASR Delta (S1-S5)
42
May 7, 2026
0.2ASR
40
Mar 30, 2026
8.3Harmful Score
40
Feb 26, 2026
3.5ASR
39
Jun 2, 2026
93.25PAIR
39
Jun 5, 2026
89F1 Score
39
Jul 2, 2026
0HRR
38
Apr 21, 2026
86.6Refusal Rate
36
May 28, 2026
95XSTest Score
36
Jun 11, 2026
0.613Safety Score
36
Feb 26, 2026
0.9Expected HFR
35
Jun 2, 2026
0.4FRR
35
Jun 9, 2026
100F1 Score
34
May 7, 2026