Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Response Harmfulness Detection on SafeRLHF

72.1F1 Score

BeaverDam

7.6224.3641.157.84May 27, 2026May 28, 2026May 29, 2026
Updated 2d ago

Evaluation Results

MethodLinks
2026.05
72.1
2026.05
70.49
2026.05
70.04
2026.05
70.04
2026.05
69.98
2026.05
69.7
2026.05
69.02
2026.05
69.02
2026.05
68.72
2026.05
68.39
2026.05
67.54
2026.05
66.6
2026.05
65.1
2026.05
65.1
2026.05
64.7
2026.05
64.7
2026.05
64.42
2026.05
64.34
2026.05
64.2
2026.05
64.2
2026.05
64.05
2026.05
63.73
2026.05
63.71
2026.05
62.76
2026.05
61.25
2026.05
60
2026.05
59.3
2026.05
59.3
2026.05
55.9
2026.05
55.9
2026.05
52.4
2026.05
51.6
2026.05
51.6
2026.05
48.4
2026.05
47.07
2026.05
47.07
2026.05
44.36
2026.05
44.36
2026.05
16.92
2026.05
16.92
2026.05
10.1