Role-Playing on Role-playing human assessment set (test)
602Win CountBaichuan2-7B+PCL
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Baichuan2-7B+PCLBackbone=Baichuan2-7B, Assessment Method=GPT-4 assessment, Enhancement=PCL (Persona-Aware Contrastive Learning)2025.03 | 602 | 92 | 306 | |
| Qwen-7B+PCLBackbone=Qwen-7B, Assessment Method=GPT-4 assessment, Enhancement=PCL (Persona-Aware Contrastive Learning)2025.03 | 582 | 181 | 237 | |
| Qwen-7B+PCLComparison Baseline=Qwen-7B2025.03 | 303 | 137 | 60 | |
| Baichuan2-7B+PCLComparison Baseline=Baichuan2-7B2025.03 | 262 | 43 | 195 |