Medical Report Summarization on MIMIC-CXR (test)
69.83ROUGE-LViTAS
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ViTASInput Modality=Multimodal2026.03 | 69.83 | 29.25 | 95.61 | — | — | |
| Flan-T5 + ViTInput Modality=Multimodal2026.03 | 47.86 | 25.87 | 64.74 | 77.93 | 51.84 | |
| OFA + ResNetInput Modality=Multimodal2026.03 | 47.75 | 25.58 | 64.8 | 76.29 | 50.96 | |
| BLOOMZ-7bInput Modality=Multimodal2026.03 | 47.48 | 25.32 | 63.61 | 74.34 | 49 | |
| BARTInput Modality=Text-Only2026.03 | 47 | — | — | 65.39 | 40.9 | |
| VG-BART (MHA)Input Modality=Multimodal, Attention Type=Multi-Head Attention2026.03 | 46.89 | — | — | 65.33 | 41.1 | |
| VG-BART (Dot)Input Modality=Multimodal, Attention Type=Dot Product2026.03 | 46.68 | — | — | 65.59 | 40.92 | |
| BioBARTInput Modality=Multimodal2026.03 | 46.15 | 22.97 | 63.43 | 75.14 | 48.04 | |
| LLaVA-MedInput Modality=Multimodal2026.03 | 44.53 | — | — | 62.98 | 38.81 | |
| LLaVA-1.5Input Modality=Multimodal2026.03 | 44.32 | — | — | 62.14 | 38.81 | |
| Qwen-VLInput Modality=Multimodal2026.03 | 44.3 | — | — | 63.08 | 38.83 | |
| BERT + CvT-21Input Modality=Multimodal2026.03 | 44.14 | 17.97 | 61.47 | 71.67 | 44.95 | |
| DeepSeek-VLInput Modality=Multimodal2026.03 | 43.12 | — | — | 62.95 | 37.79 | |
| GSumInput Modality=Text-Only2026.03 | 42.98 | — | — | 59.6 | 35.18 | |
| WGSumInput Modality=Text-Only2026.03 | 41.55 | — | — | 58.92 | 31.17 | |
| PGInput Modality=Text-Only2026.03 | 36.84 | — | — | 54.44 | 26.72 | |
| Llama-3.1-8BInput Modality=Text-Only2026.03 | 32.46 | 11.67 | 54.56 | 69.03 | 29.16 | |
| Gemma-2-9b-itInput Modality=Text-Only2026.03 | 30.71 | 10.16 | 52.95 | 68.42 | 27.44 | |
| Mistral-7BInput Modality=Text-Only2026.03 | 29.81 | 8.67 | 51.58 | 66.33 | 24.91 |