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

Harmful Refusal on HarmBench adv harm

6.4ASR

Qwen2.5-14B-IT + SELF-REDTEAM

4.0419.9735.951.83Jun 9, 2025
Updated 18d ago

Evaluation Results

MethodLinks
2025.06
6.4
2025.06
13.1
2025.06
14.4
2025.06
16.1
2025.06
17.8
2025.06
22.1
2025.06
25.9
2025.06
26.5
2025.06
27.8
2025.06
65.4