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

Input Moderation on HarmBench (test)

100F1 Score

Qwen3Guard-4B-Gen-strict

55.59267.12178.6590.179Apr 5, 2026
Updated 12d ago

Evaluation Results

MethodLinks
2026.04
100
2026.04
100
2026.04
100
2026.04
99.7
2026.04
99.4
2026.04
99.2
2026.04
99.2
2026.04
99
2026.04
98.9
2026.04
98.9
2026.04
98.7
2026.04
98.7
2026.04
98.7
2026.04
98.5
2026.04
97.2
2026.04
97.2
2026.04
96.8
2026.04
96.1
2026.04
94.8
2026.04
75.2
2026.04
60.6
2026.04
57.3