Motivational Interviewing Behavior Evaluation on AnnoMI high-quality subset
1.28R/QHuman HQ
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Human HQBackbone=Human2025.02 | 1.28 | 52.1 | 51.4 | 82.4 | 53.4 | |
| CAMIBackbone=Llama-3.1 70B2025.02 | 1.11 | 83.7 | 60.5 | 90.7 | 38.4 | |
| DIIRBackbone=Llama-3.1 70B2025.02 | 0.97 | 76.8 | 71.5 | 85.4 | 45 | |
| CAMI-TEBackbone=Llama-3.1 70B2025.02 | 0.89 | 76.6 | 61.2 | 88.9 | 40.4 | |
| Chain of Strategy (CoS)Backbone=Llama-3.1 70B2025.02 | 0.77 | 75.2 | 58.9 | 87.1 | 43.9 | |
| CAMIBackbone=GPT-4o2025.02 | 0.56 | 99 | 57.3 | 96.6 | 34.8 | |
| DIIRBackbone=GPT-4o2025.02 | 0.42 | 98.9 | 82.4 | 89.1 | 54.9 | |
| CAMI-TEBackbone=GPT-4o2025.02 | 0.37 | 97.2 | 51.1 | 94.7 | 36.1 | |
| Human LQBackbone=Human2025.02 | 0.34 | 16.2 | 14.22 | 40.5 | 59.6 | |
| BaseBackbone=Llama-3.1 70B2025.02 | 0.34 | 34.3 | 13.1 | 55.4 | 47.9 | |
| Chain of Strategy (CoS)Backbone=GPT-4o2025.02 | 0.29 | 96.5 | 49.1 | 94.5 | 38.3 | |
| BaseBackbone=GPT-4o2025.02 | 0.18 | 92.3 | 37.6 | 85 | 42.7 |