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

Adversarial Attack on Q-Bench

87.23Attack Success Rate

LAMP

28.15843.49458.8374.166Jan 29, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.01
87.23
2026.01
77.56
2026.01
77.55
2026.01
76.12
2026.01
71.45
2026.01
71.42
2026.01
70.55
2026.01
65.33
2026.01
63.29
2026.01
60.23
2026.01
60.09
2026.01
59.74
2026.01
58.41
2026.01
50.1
2026.01
49.09
2026.01
46.64
2026.01
44.76
2026.01
43.66
2026.01
43.66
2026.01
43.64
2026.01
42.64
2026.01
42.55
2026.01
42.53
2026.01
41.35
2026.01
40.87
2026.01
40.51
2026.01
40.43
2026.01
40.41
2026.01
39.33
2026.01
38.77
2026.01
38.43
2026.01
38.43
2026.01
38.13
2026.01
34.66
2026.01
33.34
2026.01
31.33
2026.01
30.43