List your company

Jailbreak Defense on HarmBench (test)

0ASRBCT
-2.9416.90536.7556.595May 27, 2026

Evaluation results

MethodLinks
BCTBase Model=GPT-OSS-20B2026.050
ACTBase Model=Qwen3-8B2026.050
BCTBase Model=Phi-4-reasoning2026.050
ACTBase Model=Phi-4-reasoning2026.051
ACTBase Model=Qwen3-1.7B2026.052.4
ACTBase Model=GPT-OSS-20B2026.054.5
ACTBase Model=Gemma-4-E4B-it2026.0511.2
BaselineBase Model=GPT-OSS-20B2026.0512.2
BCTBase Model=Gemma-4-E4B-it2026.0514.3
BCTBase Model=Qwen3-8B2026.0518.4
BaselineBase Model=Phi-4-reasoning2026.0530
BCTBase Model=Qwen3-1.7B2026.0551
BaselineBase Model=Qwen3-8B2026.0562.6
BaselineBase Model=Qwen3-1.7B2026.0563.2
BaselineBase Model=Gemma-4-E4B-it2026.0573.5