List your company

Safety Guardrail on Prompt-side safety benchmarks

88.6F1 ScoreStage3-DPO
82.25683.90385.5587.197Jul 7, 2026

Evaluation results

MethodLinks
Stage3-DPOParams=4B2026.0788.6
Stage2-SFTParams=4B2026.0786.5
Stage1-SFT-v1Params=4B2026.0786.3
Stage1-SFT-v4Params=4B2026.0785.9
Stage1-SFT-v0Params=4B2026.0785.5
Qwen3Guard-8B-GenParams=8B2026.0785.2
YuFeng-XGuard-Reason-8BParams=8B2026.0784.9
Qwen3Guard-4B-GenParams=4B2026.0784.4
Stage1-SFT-v3Params=4B2026.0784
YuFeng-XGuard-Reason-0.6BParams=0.6B2026.0783.3
Stage1-SFT-v2Params=4B2026.0783.1
Qwen3Guard-0.8B-GenParams=0.8B2026.0782.5