List your company

Harmful Query on HarmB

0.18ASR (%)NSPO
-2.105613.322228.7544.1778Dec 12, 2025

Evaluation results

MethodLinks
NSPOBackbone=Llama3-8B-Instruct2025.120.18
NSPOBackbone=Qwen2.5-7B-Instruct2025.120.5
W-DOORBackbone=Qwen2.5-7B-Instruct2025.120.51
W-DOORBackbone=Llama3-8B-Instruct2025.122.81
BFPOBackbone=Llama3-8B-Instruct2025.124.16
DPO-SBackbone=Llama3-8B-Instruct2025.125.67
MoCANBackbone=Llama3-8B-Instruct2025.126.83
Llama3-8B-InstructBackbone=Llama3-8B-Instruct2025.127.5
PeCANBackbone=Llama3-8B-Instruct2025.128.83
BFPOBackbone=Qwen2.5-7B-Instruct2025.1211.33
Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct2025.1213.83
DPO-HBackbone=Qwen2.5-7B-Instruct2025.1214.67
DPO-MixBackbone=Llama3-8B-Instruct2025.1215.58
MoCANBackbone=Qwen2.5-7B-Instruct2025.1217.83
DPO-SBackbone=Qwen2.5-7B-Instruct2025.1218.33
PeCANBackbone=Qwen2.5-7B-Instruct2025.1218.48
DPO-MixBackbone=Qwen2.5-7B-Instruct2025.1218.67
SafeRLHFBackbone=Qwen2.5-7B-Instruct2025.1232.25
SafeRLHFBackbone=Llama3-8B-Instruct2025.1241.34
DPO-HBackbone=Llama3-8B-Instruct2025.1257.32