List your company

Safety Evaluation on XSTest (Comprehensive Metrics)

92Compliance (Safe)SFT
23.8841.56559.2576.935Jun 9, 2026

Evaluation results

MethodLinks
SFTTraining configuration=SFT2026.069280.873.922.682.981.9
3-level (lin.)Training configuration=3-level (lin.)2026.0683.116.96.983.311.683.283.2
baselineTraining configuration=baseline2026.0680.619.42.481.517.481.181.1
GW-DPO (bi.)Training configuration=GW-DPO (bi.)2026.0655.644.418.4916.973.369
GW-DPO (lin.)Training configuration=GW-DPO (lin.)2026.0653.546.56.1972.575.268.9
DPOTraining configuration=DPO2026.0641.258.84.8990.570.158.2
No-ISETraining configuration=No-ISE2026.0626.573.56100063.341.9