Query Validity Verification on SA-Med2D-20M (Negative Set)
91.9AccuracyHuman Verification
Evaluation Results
| Method | Links | |
|---|---|---|
| Human Verification2026.04 | 91.9 | |
| MedVeriSegReasoning Module=GPT-4o2026.04 | 85.8 | |
| MedVeriSeg w/o GPT-4oReasoning Module=None2026.04 | 83.6 | |
| Qwen2.5-VL-7B2026.04 | 82.8 | |
| LLaVA-Med2026.04 | 82.4 | |
| GPT-4o2026.04 | 77.8 | |
| DeepSeek-Janus-Pro2026.04 | 66.9 | |
| Random Verification2026.04 | 46.5 |