List your company

Multi-turn Conversation Evaluation on MT-Bench 1.0 (test)

8.538GPT-4 ScoreDAR
8.325848.380928.4368.49108Feb 12, 2026

Evaluation results

MethodLinks
DARResponse Length=13582026.028.5387.931
GRPOResponse Length=15592026.028.4257.856
RLOOResponse Length=15802026.028.4097.893
Iter-SFTResponse Length=13432026.028.3787.838
Qwen2-7B-Instruct (π0)Response Length=13402026.028.3347.769