List your company

Alignment on HH IDN 40%

68Win RateUDPO
45.1251.065762.94Jul 3, 2026

Evaluation results

MethodLinks
UDPOBase Model=Qwen-3-1.7B2026.0768
α-UDPOBase Model=Qwen-3-1.7B2026.0767.6
rDPOBase Model=Qwen-3-1.7B2026.0763.7
DPOBase Model=Qwen-3-1.7B2026.0756.4
α-URMBase Model=Qwen-3-1.7B2026.0755.2
URMBase Model=Qwen-3-1.7B2026.0754.6
rRMBase Model=Qwen-3-1.7B2026.0748.2
RMBase Model=Qwen-3-1.7B2026.0746