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

Prompt Harmfulness Detection on AegisSafety (test)

99.5F1 Score

LLaMA Guard 3

3.788828.636953.48578.3331Feb 28, 2025May 14, 2025Jul 29, 2025Oct 13, 2025Dec 28, 2025Mar 14, 2026May 29, 2026
Updated 2d ago

Evaluation Results

MethodLinks
2026.05
99.5
2026.05
91.39
2026.05
91.39
2026.05
91.02
2026.05
90.58
2026.05
90.18
2026.05
90.18
2026.05
89.69
2026.05
89.45
2026.05
89.4
2026.05
89.34
2026.05
88.24
2026.05
86.58
2026.05
86.32
2026.05
84.8
2026.05
84.8
2026.05
83.36
2026.05
83.15
2026.05
82.9
2026.05
82.9
2026.05
82.88
2026.05
81.96
2026.05
81.07
2026.05
80.52
2026.05
80.5
2026.05
80.23
2026.05
78.62
2026.05
77.63
2026.05
77.63
2025.02
74.8
2025.02
74.1
2026.05
74.1
2025.02
74
2026.05
71.8
2026.05
71.8
2026.05
71.39
2025.02
31.9
2026.05
31.9
2025.02
7.5
2026.05
7.47
2026.05
7.47