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

LLM Jailbreaking on HarmBench text (test N = 320)

93.75ASR-M

PEO

36.872451.638766.40581.1713Apr 27, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
93.7550
2026.04
92.1916.88
2026.04
90.9449.69
2026.04
86.2521.88
2026.04
82.8145.62
2026.04
8033.44
2026.04
70.3137.5
2026.04
69.3843.75
2026.04
68.4416.88
2026.04
66.5639.38
2026.04
64.6919.38
2026.04
58.4426.25
2026.04
53.4412.5
2026.04
47.1924.06
2026.04
406.88
2026.04
39.068.12