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

Visual Compliance Verification on UnsafeBench

76Unsafe F1

CompAgent

33.3644.4355.566.57Oct 31, 2025
Updated 27d ago

Evaluation Results

MethodLinks
2025.10
7682708781
2025.10
7170717876
2025.10
6862747372
2025.10
68---78
2025.10
6679567874
2025.10
6380527773
2025.10
6378537673
2025.10
5945865453
2025.10
5679437469
2025.10
5587407569
2025.10
5479417368
2025.10
5487357468
2025.10
5465467166
2025.10
50----
2025.10
3588226957