Safety Evaluation
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
100Safe Response Rate
4
Feb 26, 2026
100Safe Response Rate
4
Feb 26, 2026
100Safety Score
4
Feb 26, 2026
100Safety Score
4
Feb 26, 2026
100Safety Score
4
Feb 26, 2026
96.6Safety Score
4
Feb 26, 2026
1Safety Score
4
Feb 26, 2026
0.997Safety Score
4
Feb 26, 2026
99.3Safety Score
4
Feb 26, 2026
92.6Hate (Non-Violent)
4
Feb 26, 2026
100Hate Speech Aggregate Accuracy
4
Feb 26, 2026
88.3Non-Violent Hate Score
4
Feb 26, 2026
73.6Generation Quality (Std Prefix)
4
Feb 26, 2026
100StrongReject Score
4
Feb 26, 2026
55.67Chinese Accuracy
4
Feb 26, 2026
0.534Score (Chinese)
4
Feb 26, 2026
86.9F1 Score
4
Feb 26, 2026
43.67HADES Score
4
Feb 26, 2026
34.62Model-level Safety Score
3
May 29, 2026
0.39Unsafe Rate
3
May 22, 2026
0.639Harm Score Delta (Δ)
3
May 15, 2026
1.3ASR
3
Apr 14, 2026
1ASR
3
Apr 14, 2026
0.02ASR
3
Apr 14, 2026
0Attack Success Rate (ASR)
3
Apr 13, 2026