Note Generation on MM-HEALTHCN Video
97Relevance (R)EVONOTE-NM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EVONOTE-NMBackbone=MedGemma-27B, Memory=removed retrieval and update2026.06 | 97 | 82 | 76 | 85.5 | |
| EVONOTEBackbone=MedGemma-27B2026.06 | 97 | 90.5 | 84.25 | 90.13 | |
| ReMemBackbone=MedGemma-27B2026.06 | 95.25 | 81 | 70.25 | 74 | |
| EVONOTE-NM-NABackbone=MedGemma-27B, Memory=removed retrieval and update, Claim Analyzer=removed2026.06 | 93.5 | 73.5 | 67.5 | 77.63 | |
| Grok-1.52026.06 | 91 | 71.5 | 67 | 76.38 | |
| DeepResearch2026.06 | 90 | 75.75 | 73.25 | 76.38 | |
| ExpRAGBackbone=MedGemma-27B2026.06 | 89.5 | 74 | 68 | 80.75 | |
| EVONOTE-FMBackbone=MedGemma-27B, Memory=frozen warm-up2026.06 | 88.5 | 78.25 | 71 | 82.75 | |
| CrowdNotes+Backbone=MedGemma-27B2026.06 | 87.75 | 71 | 60.5 | 67.75 | |
| Human Baseline2026.06 | 76.75 | 54 | 39.25 | — | |
| Gemini-1.5-Flash2026.06 | 58 | 47.75 | 46 | 72.88 | |
| GPT-4o2026.06 | 32.75 | 27.5 | 25.25 | 65.13 |