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

Toxicity Defense on LLaVA v1.5

22.35Toxicity Score

NullSteer

19.822436.883753.94571.0063Mar 23, 2026
Updated 25d ago

Evaluation Results

MethodLinks
2026.03
22.35
2026.03
25.1
2026.03
31.82
2026.03
32.55
2026.03
34.76
2026.03
36.02
2026.03
39.32
2026.03
43.13
2026.03
50.21
2026.03
52.61
2026.03
54.9
2026.03
55.48
2026.03
60.36
2026.03
62.46
2026.03
66.72
2026.03
66.81
2026.03
66.82
2026.03
71.32
2026.03
73.14
2026.03
73.76
2026.03
73.77
2026.03
75
2026.03
75
2026.03
75
2026.03
75
2026.03
75.76
2026.03
77.6
2026.03
77.77
2026.03
80.93
2026.03
83.7
2026.03
83.92
2026.03
83.97
2026.03
84.19
2026.03
84.4
2026.03
85.44
2026.03
85.54