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

Harmful Refusal on OR-Bench vani harm

98.5RTA

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

Updated 18d ago

Evaluation Results

Method	Links
Qwen2.5-14B-IT + SELF-REDTEAM 2025.06		98.5
Qwen2.5-3B-IT + SELF-REDTEAM 2025.06		97.2
Qwen2.5-7B-IT + SELF-REDTEAM 2025.06		96.8
Llama3.1-8B-IT + SELF-REDTEAM 2025.06		94.2
Qwen2.5-3B-IT 2025.06		93
Qwen2.5-14B-IT 2025.06		89.3
Qwen2.5-7B-IT 2025.06		87.9
Llama3.1-8B-IT 2025.06		86.4
Llama3.1-8B-IT-AB + SELF-REDTEAM 2025.06		84.6
Llama3.1-8B-IT-AB 2025.06		1.4