Multi-turn role-play on MSD
83.2Success Rate (SR)EmoLLM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EmoLLMBackbone=Qwen3-8B, Method Category=Ours2026.03 | 83.2 | 2.86 | 4.17 | 3.71 | |
| PERMBackbone=Qwen3-8B, Method Category=RL Methods2026.03 | 82.1 | 2.74 | 4.2 | 3.64 | |
| ICL-AIFBackbone=Qwen3-8B, Method Category=Prompting Methods2026.03 | 81.5 | 2.81 | 4.16 | 3.62 | |
| SFTBackbone=Qwen3-8B, Method Category=SFT Methods2026.03 | 81.2 | 2.74 | 3.97 | 3.37 | |
| RLVERBackbone=Qwen3-8B, Method Category=RL Methods2026.03 | 81.2 | 2.88 | 4.18 | 3.61 | |
| ARG-PromptBackbone=Qwen3-8B, Method Category=Prompting Methods2026.03 | 81 | 2.88 | 4.14 | 3.58 | |
| RLFF-ESCBackbone=Qwen3-8B, Method Category=RL Methods2026.03 | 80.6 | 2.96 | 4.15 | 3.55 | |
| PPDPBackbone=Qwen3-8B, Method Category=SFT Methods2026.03 | 80.4 | 2.93 | 4.11 | 3.54 | |
| Best-of-NBackbone=Qwen3-8B, Method Category=Prompting Methods2026.03 | 79.6 | 3.12 | 4.08 | 3.51 | |
| ESCoTBackbone=Qwen3-8B, Method Category=SFT Methods2026.03 | 79.4 | 3.06 | 4.18 | 3.47 | |
| Ask-an-ExpertBackbone=Qwen3-8B, Method Category=Prompting Methods2026.03 | 79.1 | 3.15 | 4.05 | 3.46 | |
| Qwen3-8BBackbone=Qwen3-8B, Method Category=Base Model2026.03 | 77.6 | 3.27 | 4.13 | 3.41 |