Harmful Content Detection
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
89.5AUPRC
61
Jun 25, 2026
75.6AUPRC
49
May 4, 2026
82.91Accuracy
15
Feb 26, 2026
83.6F1 Score
12
Jun 26, 2026
18.6ASR
12
May 12, 2026
85.59Accuracy
10
Feb 26, 2026
48.57Accuracy
8
Apr 21, 2026
43.96Accuracy
8
Apr 21, 2026
90.19Accuracy
8
Apr 21, 2026
96.4Accuracy
8
Apr 21, 2026
97.24Accuracy
8
Apr 21, 2026
98.47Accuracy
8
Apr 21, 2026
97.09Accuracy
8
Apr 21, 2026
99.06Accuracy
8
Apr 21, 2026
2.75FNR
7
Apr 3, 2026
0.793AUROC
6
May 13, 2026
92.6Safe Rate
4
Jul 7, 2026
16.24Accuracy
4
Apr 21, 2026
26.04Accuracy
4
Apr 21, 2026
19.88Accuracy
4
Apr 21, 2026
60.13Accuracy
4
Apr 21, 2026
69.18Accuracy
4
Apr 21, 2026
64.84Accuracy
4
Apr 21, 2026
19.88Accuracy
4
Apr 21, 2026
65.55Accuracy
4
Apr 21, 2026