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

LLM Safety on Safety Evaluation Set

1.66Harmful Response Rate

Probing Vector

1.39923.15964.926.6804Feb 11, 2026
Updated 3mo ago

Evaluation Results

MethodLinks
2026.02
1.66
2026.02
2.51
2026.02
2.57
2026.02
2.86
2026.02
3.1
2026.02
3.3
2026.02
3.48
2026.02
3.61
2026.02
3.73
2026.02
3.75
2026.02
3.9
2026.02
4.63
2026.02
4.96
2026.02
5.13
2026.02
5.16
2026.02
5.78
2026.02
6.33
2026.02
6.38
2026.02
6.67
2026.02
6.71
2026.02
6.78
2026.02
6.89
2026.02
7.63
2026.02
7.68
2026.02
8.18