Role-playing evaluation on DynSess-Eval
4.35Average Performance (Auto)Claude Sonnet 4.6
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Claude Sonnet 4.6Domain=General Models2026.05 | 4.35 | 3.24 | 3.76 | 3.12 | 4.55 | 3.21 | 4.88 | 3.37 | 4.24 | 3.27 | |
| GPT-5.4Domain=General Models2026.05 | 4.29 | 3.31 | 3.8 | 3.23 | 4.27 | 3.34 | 4.88 | 3.38 | 4.29 | 3.28 | |
| Deepseek v3.2Domain=General Models, Params=685B2026.05 | 4.2 | 3.21 | 3.7 | 3.11 | 4.25 | 3.17 | 4.73 | 3.34 | 4.12 | 3.22 | |
| Gemini-3-proDomain=General Models2026.05 | 4.12 | 3.17 | 3.76 | 3.05 | 3.67 | 3.06 | 4.83 | 3.36 | 4.23 | 3.2 | |
| Doubao-1.5-pro-characterDomain=Character Models2026.05 | 3.96 | 3.38 | 3.8 | 3.29 | 3.66 | 3.42 | 4.5 | 3.47 | 3.97 | 3.35 | |
| Qwen-plus-characterDomain=Character Models2026.05 | 3.66 | 3.01 | 3.57 | 2.86 | 3.46 | 2.88 | 4.02 | 3.22 | 3.59 | 3.06 | |
| MiniMax-M2-HERDomain=Character Models, Params=32B2026.05 | 3.23 | 2.99 | 3.18 | 2.95 | 2.89 | 2.91 | 3.76 | 3.13 | 3.11 | 2.97 | |
| CoserDomain=Character Models, Params=70B2026.05 | 2.87 | 2.96 | 2.61 | 2.89 | 2.37 | 2.9 | 3.34 | 3.09 | 3.12 | 2.98 |