Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Harmful prompt detection on HarmB
Loading...
100
F1 Score
MLPM
67.8016
76.1608
84.52
92.8792
Feb 22, 2025
F1 Score
Updated 1d ago
Evaluation Results
Method
Method
Links
F1 Score
MLPM
Backbone=Qwen3-8B-Inst...
2025.02
100
MLPM
Backbone=Llama-8B-Inst...
2025.02
99.58
Abdelnabi et al.
Backbone=Qwen3-8B-Inst...
2025.02
99.37
WildGuard
Methodology=Guard Model
2025.02
99.37
MLPM
Backbone=Mistral-7B-In...
2025.02
99.16
LlamaGuard3
Methodology=Guard Model
2025.02
98.94
MLPM
Backbone=OLMo2-7B-Inst...
2025.02
98.51
Abdelnabi et al.
Backbone=Mistral-7B-In...
2025.02
97.86
Ayub & Majumdar
Backbone=Llama-8B-Inst...
2025.02
96.98
Ayub & Majumdar
Backbone=OLMo2-7B-Inst...
2025.02
96.54
Abdelnabi et al.
Backbone=Llama-8B-Inst...
2025.02
95.18
Abdelnabi et al.
Backbone=OLMo2-7B-Inst...
2025.02
93.3
Ayub & Majumdar
Backbone=Mistral-7B-In...
2025.02
90.87
Ayub & Majumdar
Backbone=Qwen3-8B-Inst...
2025.02
90.62
GraniteGuardian-3-1-8B
Methodology=Guard Model
2025.02
79.9
Aegis-Guard-D
Methodology=Guard Model
2025.02
70.46
ShieldGemma-9B
Methodology=Guard Model
2025.02
69.04
Feedback
Search any
task
Search any
task