Query Validity Verification on SA-Med2D-20M (All Item Set)
90.2AccuracyHuman Verification
Evaluation Results
| Method | Links | |
|---|---|---|
| Human Verification2026.04 | 90.2 | |
| MedVeriSegReasoning Module=GPT-4o2026.04 | 85.3 | |
| MedVeriSeg w/o GPT-4oReasoning Module=None2026.04 | 83.1 | |
| LLaVA-Med2026.04 | 82.2 | |
| Qwen2.5-VL-7B2026.04 | 82 | |
| GPT-4o2026.04 | 76.3 | |
| DeepSeek-Janus-Pro2026.04 | 63.7 | |
| Random Verification2026.04 | 47.2 |