Breast Ultrasound Diagnosis on BUSI (in-domain test)
0.835AUCQwen2.5-3B-COT-SFT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-3B-COT-SFTBackbone=Qwen2.5-3B, Training Strategy=SFT on template-based reasoning data, Inference Workflow=Intermediate attribute evidence2026.03 | 0.835 | 83.3 | 46 | 0.245 | |
| UltrasoundAgentsBackbone=Qwen2.5-3B, Training Strategy=Three-stage RL (GRPO) and SFT, Inference Workflow=ROI -> attribute -> diagnosis evidence-chain2026.03 | 0.784 | 83.3 | 54.2 | 0.244 | |
| Think-with-ImageBackbone=Qwen2.5-3B, Training Strategy=DeepEyes-style single-agent workflow, Inference Workflow=ROI localization -> crop-and-zoom -> re-analyze2026.03 | 0.528 | 66 | 8 | 0.019 | |
| Qwen2.5-3B-Zero-ShotBackbone=Qwen2.5-3B, Training Strategy=Zero-shot, Inference Workflow=Directly prompts on full image2026.03 | 0.493 | 57.3 | 15.6 | 0.066 |