List your company

Role-playing on DynSess-Eval Human Evaluation 1.0 (test)

3.38Average ScoreDoubao-1.5-pro-character
2.94323.05663.173.2834May 28, 2026

Evaluation results

MethodLinks
Doubao-1.5-pro-characterDomains=Character Models, Params=∼200B12026.053.383.293.423.473.35
DynSess-Character-32B (DSPO)Domains=Ours, Params=32B2026.053.373.233.343.563.34
DynSess-Character-32B (GSRPO)Domains=Ours, Params=32B2026.053.353.393.253.463.31
GPT-5.4Domains=General Models, Params=-2026.053.313.233.343.383.28
Claude Sonnet 4.6Domains=General Models, Params=-2026.053.243.123.213.373.27
Deepseek v3.2Domains=General Models, Params=685B2026.053.213.113.173.343.22
Gemini-3-proDomains=General Models, Params=-2026.053.173.053.063.363.2
Qwen-plus-characterDomains=Character Models, Params=-2026.053.012.862.883.223.06
MiniMax-M2-HERDomains=Character Models, Params=32B2026.052.992.952.913.132.97
CoserDomains=Character Models, Params=70B2026.052.962.892.93.092.98