Close-Ended Visual Question Answering on CXR
79AccuracyHR1.5
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HR1.52026.07 | 79 | — | — | |
| CheXOne-Reasoning2026.07 | 77.6 | — | — | |
| GPT-5.42026.07 | 77.5 | — | — | |
| HR1.5+2026.07 | 76.2 | — | — | |
| HR12026.07 | 73.1 | — | — | |
| CheXOne-Instruct2026.07 | 72.7 | — | — | |
| Gemini-3-Flash-Preview2026.07 | 67.9 | — | — | |
| MedGemma-27B2026.07 | 65.6 | — | — | |
| MedGemma-4B2026.07 | 64.6 | — | — | |
| Claude Opus 4.72026.07 | 59.5 | — | — | |
| LLaMa-4-Scout2026.07 | 44.5 | — | — | |
| CheXagent2026.07 | 34.7 | — | — | |
| AnatomiX2026.01 | — | 89 | 95 | |
| AnatomiX-IpAblation=Image embeddings only, excluding anatomical tokens and retrieved sentences2026.01 | — | 61 | 87 | |
| AnatomiX-O^AAblation=Image embeddings plus anatomical tokens, excluding ybox and retrieved sentences2026.01 | — | 48 | 79 | |
| CheXagent2026.01 | — | 90 | 97 | |
| MAIRA-22026.01 | — | 10 | 81 | |
| MedGemma2026.01 | — | 2 | 38 | |
| Radialog2026.01 | — | 3 | 92 | |
| RadVLM2026.01 | — | 23 | 67 |