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

Harmful prompt detection on OpenAI

81.3F1 Score

Stage1-SFT-v2

63.921668.433372.94577.4567Feb 22, 2025May 16, 2025Aug 7, 2025Oct 30, 2025Jan 21, 2026Apr 14, 2026Jul 7, 2026
Updated 17d ago

Evaluation Results

MethodLinks
2026.07
81.3
2026.07
80.3
2026.07
79.8
2025.02
79.11
2026.07
79
2025.02
77.63
2025.02
77.63
2026.07
76.8
2025.02
76.44
2026.07
75.8
2025.02
75.69
2025.02
75.19
2026.07
74.7
2025.02
74.56
2025.02
74.21
2026.07
74
2025.02
72.85
2026.07
72.8
2025.02
72.35
2025.02
72.28
2025.02
70.68
2026.07
68.5
2025.02
68.45
2026.07
68.4
2025.02
67.99
2025.02
66.95
2025.02
66.6
2026.07
66.2
2025.02
64.59