Utterance-level Motivational Interviewing Coding on Motivational Interviewing (MI) Coding Dataset
52.56AccuracyMM-SC
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MM-SCPrompting Strategy=Multimodal Self-Consistency (MM-SC), Backbone Model=Qwen3-Omni-30B-A3B-Instruct, Temperature=1.0, Top-p=0.52026.05 | 52.56 | 54.03 | 47.45 | 46.4 | |
| COT (Baseline)Prompting Strategy=Chain-of-Thought (COT) Prompting, Backbone Model=Qwen3-Omni-30B-A3B-Instruct, Temperature=1.0, Top-p=0.52026.05 | 49.89 | 50.32 | 44.03 | 41.97 | |
| Direct (Baseline)Prompting Strategy=Direct Prompting, Backbone Model=Qwen3-Omni-30B-A3B-Instruct, Temperature=1.0, Top-p=0.52026.05 | 46.22 | 47.22 | 45.05 | 42.48 |