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

Safety Alignment on WildJailbreak

15,768.31Trainable parameters (M)

Full FT

-630.59723,626.81147,884.2212,141.6286Feb 18, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.02
15,768.31
2026.02
14,770.03
2026.02
14,660
2026.02
8,537.68
2026.02
7,616.48
2026.02
7,615.61
2026.02
4,022.46
2026.02
3,836
2026.02
3,212.75
2026.02
2,506.17
2026.02
1,235.81
2026.02
68.4
2026.02
51.6
2026.02
37.5
2026.02
31.52
2026.02
30.27
2026.02
29.28
2026.02
19.26
2026.02
18.81
2026.02
18.23
2026.02
14.16
2026.02
13.36
2026.02
8.45
2026.02
6.63
2026.02
5.04
2026.02
4.07
2026.02
3.27
2026.02
1.8
2026.02
1.62
2026.02
1.54
2026.02
1.01
2026.02
0.89
2026.02
0.87
2026.02
0.62
2026.02
0.45
2026.02
0.45
2026.02
0.44
2026.02
0.42
2026.02
0.38
2026.02
0.36
2026.02
0.36
2026.02
0.21
2026.02
0.17
2026.02
0.13