List your company

General Chat Evaluation on Arena-Hard

84Win RateVRM-PPO
42.29653.12363.9574.777Mar 5, 2026

Evaluation results

MethodLinks
VRM-PPOBackbone=Qwen 3-8B2026.038482
DPOBackbone=Qwen 3-8B2026.0370.871.9
KTOBackbone=Qwen 3-8B2026.0369.570.1
VRM-PPOBackbone=Qwen2.5-7B2026.0365.557.2
IPOBackbone=Qwen 3-8B2026.0365.554.9
WPOBackbone=Qwen2.5-7B2026.036253
SIMPOBackbone=Qwen 3-8B2026.0361.956.6
SELECTIVE DPOBackbone=Qwen 3-8B2026.0360.451.8
PPOBackbone=Qwen 3-8B2026.0360.471.9
WPOBackbone=Qwen 3-8B2026.0356.253.7
PPOBackbone=Qwen2.5-7B2026.0355.348.9
IPOBackbone=Qwen2.5-7B2026.0354.648.3
SIMPOBackbone=Qwen2.5-7B2026.0354.648.8
SELECTIVE DPOBackbone=Qwen2.5-7B2026.0351.748.2
DPOBackbone=Qwen2.5-7B2026.034947.2
KTOBackbone=Qwen2.5-7B2026.0343.944.1