List your company

Alignment on HH (IDN 20%)

78.2Win Rateα-UDPO
55.94461.72267.573.278Jul 3, 2026

Evaluation results

MethodLinks
α-UDPOBase Model=Qwen-3-1.7B2026.0778.2
UDPOBase Model=Qwen-3-1.7B2026.0777.9
rDPOBase Model=Qwen-3-1.7B2026.0773.3
DPOBase Model=Qwen-3-1.7B2026.0768.7
α-URMBase Model=Qwen-3-1.7B2026.0760.2
URMBase Model=Qwen-3-1.7B2026.0759.5
rRMBase Model=Qwen-3-1.7B2026.0758
RMBase Model=Qwen-3-1.7B2026.0756.8