List your company

Chatbot Evaluation on ArenaHard

13.88Win RateAttentionPO
1.79524.93268.0711.2074May 21, 2026

Evaluation results

MethodLinks
AttentionPOReference Model=Mistral-7B-Base-SFT2026.0513.88
DPOReference Model=Mistral-7B-Base-SFT2026.0512.45
Mistral-7B-Base-SFTReference Model=Mistral-7B-Base-SFT2026.052.26