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

Backdoor Defense on Refusal behavior dataset

80.3CACC (BadNet)

Fine-mixing

22.0637.1852.367.42Jan 7, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.01
80.30.577.16177.532.678.42.3
2026.01
79.80.579.83.279.81.479.80.9
2026.01
77.30.973.42.375.70.577.10.5
2026.01
76.624.377.174.867.980.370.120.6
2026.01
75.74.174.87874.368.379.42.3
2026.01
73.47.877.572.566.571.673.112.4
2026.01
39.4038.10.937.2036.20
2026.01
391.4391.4391.8391.8
2026.01
32.119.33377.131.246.827.119.3
2026.01
29.440.433.98333.970.626.634.9
2026.01
25.251.433.979.432.679.42251.8
2026.01
24.35535.387.628.478.918.344