List your company

Safety Evaluation on Harmful Prompts (ASR Metrics)

2ASR (Raw)No Intervention
-1.6823.164872.84May 28, 2026

Evaluation results

MethodLinks
No InterventionBackbone=DeepSeek2026.052——
No InterventionBackbone=Qwen32026.052——
No InterventionBackbone=Phi2026.052.5——
No InterventionBackbone=OLMoE2026.054——
No InterventionBackbone=GPT-oss2026.055——
GCGBackbone=GPT-oss2026.056.551.5
GCGBackbone=Qwen32026.05910
GCGBackbone=OLMoE2026.051410.52.5
GCGBackbone=DeepSeek2026.051520
GCGBackbone=Phi2026.05180.50
RASETBackbone=Phi2026.056155.529.5
RASETBackbone=GPT-oss2026.0567.56139.5
RASETBackbone=OLMoE2026.057875.550.5
RASETBackbone=DeepSeek2026.0592.59061.5
RASETBackbone=Qwen32026.05948871.5