Safety Evaluation
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
97.2HEx-PHI Score
162
Mar 13, 2026
0ASR
153
Jun 9, 2026
2Harmfulness
140
Apr 24, 2026
0.06Harmbench Score
127
May 29, 2026
100Safety Score
117
Jun 1, 2026
7.9Harmful Score
110
May 26, 2026
1.8Attack Success Rate (ASR)
107
Jun 30, 2026
0Average ASR
98
May 28, 2026
0HRR
96
Feb 26, 2026
99.09Safety Score
90
Feb 26, 2026
0.015ASR
90
Jun 16, 2026
9Attack Success Rate
87
Jun 2, 2026
0False Refusal Rate (FR)
84
Jun 26, 2026
5Harmfulness Score
84
Apr 24, 2026
0.79Harmfulness Score
84
Feb 26, 2026
4FC
78
May 7, 2026
0Attack Success Rate
77
May 12, 2026
100Safety
77
Jun 5, 2026
99Safe Response Rate
66
Mar 13, 2026
0.55Safety Score
60
Feb 26, 2026
0.4ASR
57
Mar 31, 2026
8.59AdvBench Score
56
Jun 2, 2026
1.12ASR (Llama-Guard-3-8B)
56
May 7, 2026
7.5ASR
56
May 7, 2026
4.44ASR
56
May 7, 2026