List your company

Safety Alignment Evaluation on WildJailbreak

98.6Safety RateStair-DPO
41.19256.0967185.904May 30, 2026

Evaluation results

MethodLinks
Stair-DPOArchitecture=Qwen3-30B-A3B2026.0598.6
MESAArchitecture=Qwen3-30B-A3B2026.0597.65
SafeXArchitecture=Qwen3-30B-A3B2026.0596.35
SFTArchitecture=Qwen3-30B-A3B2026.0595
GRPOArchitecture=Qwen3-30B-A3B2026.0594.4
MESAArchitecture=DeepSeek-v2-Lite2026.0590.9
Stair-DPOArchitecture=DeepSeek-v2-Lite2026.0583.6
Stair-SFTArchitecture=DeepSeek-v2-Lite2026.0577.9
SFTArchitecture=DeepSeek-v2-Lite2026.0577.7
Base(instruct)Architecture=Qwen3-30B-A3B2026.0575.3
SafeXArchitecture=DeepSeek-v2-Lite2026.0564
Stair-SFTArchitecture=Qwen3-30B-A3B2026.0562.65
GRPOArchitecture=DeepSeek-v2-Lite2026.0544.1
Base(chat)Architecture=DeepSeek-v2-Lite2026.0543.4