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

Targeted Refusal on LLaMA2-13B-Chat (test)

100BadNets

No Defense

12.79635.435558.07580.7145Mar 13, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.03
10074.8683.0796.5384.2884.75
2026.03
98.9832.5750.693.8766.1568.43
2026.03
98.5634.2652.3295.9468.3369.88
2026.03
98.5475.6381.2697.2486.4587.82
2026.03
98.3274.2181.2595.3787.2487.28
2026.03
93.846.7854.8695.8384.5275.16
2026.03
93.2135.6248.3784.878.6267.92
2026.03
90.172.5483.4389.5278.4282.8
2026.03
83.6534.8662.7882.2545.3361.77
2026.03
30.1624.3226.6432.3418.8626.46
2026.03
16.159.8312.6518.237.8212.94