Counterfactual Medical Image Generation on MIMIC-EDIT-INSTRUCTION (test)
84.66Pathology ScoreReal Images (GT)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Real Images (GT)2026.02 | 84.66 | 99.56 | 88.83 | 90.99 | — | — | |
| RoentGen2026.02 | 84.53 | 53.15 | 18.6 | 67.74 | 48.92 | 35.19 | |
| InstructX2XReasoning Step-by-step Editing (RSE)=false2026.02 | 83.54 | 96.15 | 82.04 | 88.87 | 14.99 | 1.92 | |
| InstructX2XReasoning Step-by-step Editing (RSE)=true2026.02 | 83.33 | 97.65 | 82.84 | 89.35 | 7.88 | 2.64 | |
| RadEdit2026.02 | 82.8 | 80.33 | 48.24 | 80.14 | 38.85 | 26.31 | |
| BiomedJourney2026.02 | 81.27 | 79.07 | 64.37 | 80.31 | 20.45 | 12.46 | |
| LLM-CXR2026.02 | 55.82 | 48.62 | 0.1 | 52.06 | 60.03 | 51.42 |