Multi-turn dialogue evaluation on BotChat
86.3Success Rate (N=16)ISM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ISMBackbone=Qwen2.5-14B2024.08 | 86.3 | 94 | |
| SFTBackbone=Qwen2.5-14B2024.08 | 85.6 | 93.4 | |
| ISMBackbone=LLaMA3.1-8B2024.08 | 79.2 | 93.6 | |
| ISMBackbone=Qwen2.5-7B2024.08 | 78.2 | 92 | |
| SFTBackbone=LLaMA3.1-8B2024.08 | 77.3 | 90.5 | |
| SFTBackbone=Qwen2.5-7B2024.08 | 71.8 | 89.4 | |
| DirectBackbone=LLaMA3.1-8B2024.08 | 67.3 | 89.8 | |
| DirectBackbone=Qwen2.5-7B2024.08 | 57.6 | 90.1 | |
| DirectBackbone=Qwen2.5-14B2024.08 | 57 | 87.2 |