List your company

Dialogue on HH (Anthropic Helpful and Harmless)

82.5Win Rate (0% Flip)α-URM
62.63667.79372.9578.107Jul 3, 2026

Evaluation results

MethodLinks
α-URMModels=Llama-3.2-3B2026.0782.581.474.379.9
URMModels=Llama-3.2-3B2026.0781.478.972.679.4
cRMModels=Llama-3.2-3B2026.0780.776.170.175.7
RMModels=Llama-3.2-3B2026.078078.369.775.5
rRMModels=Llama-3.2-3B2026.0779.978.372.475.7
URMModels=Qwen-3-1.7B2026.0767.661.858.270.1
α-URMModels=Qwen-3-1.7B2026.0766.961.657.969.6
rRMModels=Qwen-3-1.7B2026.0765.762.552.165.9
cRMModels=Qwen-3-1.7B2026.0764.25646.265
RMModels=Qwen-3-1.7B2026.0763.455.543.963.9