Query Validity Verification on SA-Med2D-20M (Positive Set)
85.6AccuracyHuman Verification
Evaluation results
| Method | Links | |
|---|---|---|
| Human Verification2026.04 | 85.6 | |
| MedVeriSegReasoning Module=GPT-4o2026.04 | 83.8 | |
| LLaVA-Med2026.04 | 81.7 | |
| MedVeriSeg w/o GPT-4oReasoning Module=None2026.04 | 81.7 | |
| Qwen2.5-VL-7B2026.04 | 79.9 | |
| GPT-4o2026.04 | 72.3 | |
| DeepSeek-Janus-Pro2026.04 | 55 | |
| Random Verification2026.04 | 49.3 |