Conversational Ability on CharacterEval
3.612FluencyGPT-4+PCL*
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-4+PCL*Backbone=GPT-4, Evaluation Mode=PCL* (Chain of Persona Prompting)2025.03 | 3.612 | 3.796 | 3.551 | 3.653 | |
| Qwen-7B+PCLBackbone=Qwen-7B, Evaluation Mode=PCL2025.03 | 3.41 | 3.736 | 3.224 | 3.457 | |
| GPT-4+ICLBackbone=GPT-4, Evaluation Mode=ICL2025.03 | 3.332 | 3.669 | 3.343 | 3.448 | |
| Baichuan2-7B+PCLBackbone=Baichuan2-7B, Evaluation Mode=PCL2025.03 | 3.263 | 3.456 | 3.288 | 3.336 | |
| Qwen-7B+SFT+ICLBackbone=Qwen-7B, Evaluation Mode=SFT+ICL2025.03 | 3.251 | 3.598 | 3.236 | 3.362 | |
| Qwen-7B+ICLBackbone=Qwen-7B, Evaluation Mode=ICL2025.03 | 3.187 | 3.564 | 3.229 | 3.327 | |
| Baichuan2-7B+SFT+ICLBackbone=Baichuan2-7B, Evaluation Mode=SFT+ICL2025.03 | 3.128 | 3.441 | 3.279 | 3.282 | |
| CharacterGLM-6B+PCLBackbone=CharacterGLM-6B, Evaluation Mode=PCL2025.03 | 3.102 | 3.164 | 2.911 | 3.059 | |
| Baichuan2-7B+ICLBackbone=Baichuan2-7B, Evaluation Mode=ICL2025.03 | 3.051 | 3.394 | 3.327 | 3.257 | |
| GPT-3.5+PCL*Backbone=GPT-3.5, Evaluation Mode=PCL* (Chain of Persona Prompting)2025.03 | 2.959 | 3.27 | 2.719 | 2.983 | |
| CharacterGLM-6B+SFT+ICLBackbone=CharacterGLM-6B, Evaluation Mode=SFT+ICL2025.03 | 2.784 | 2.765 | 2.717 | 2.755 | |
| CharacterGLM-6B+ICLBackbone=CharacterGLM-6B, Evaluation Mode=ICL2025.03 | 2.559 | 2.668 | 2.504 | 2.577 | |
| GPT-3.5+ICLBackbone=GPT-3.5, Evaluation Mode=ICL2025.03 | 2.548 | 2.872 | 2.471 | 2.63 |