Report Generation on M3TAVR (test)
42.6ROUGE-LR-CGA + InternVL3.5-8B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| R-CGA + InternVL3.5-8BFramework=TAVR-VLM, Backbone=InternVL3.5-8B2026.06 | 42.6 | 93.6 | 80.1 | |
| R-CGA + Qwen3-VL-8BFramework=TAVR-VLM, Backbone=Qwen3-VL-8B2026.06 | 42.2 | 92.8 | 78.9 | |
| Gemini-3 ProParadigm=Closed-Source Frontier Models2026.06 | 41.8 | 91.4 | 77.2 | |
| Claude-4 OpusParadigm=Closed-Source Frontier Models2026.06 | 41.5 | 90.8 | 76.4 | |
| GPT-4oParadigm=Closed-Source Frontier Models2026.06 | 41.2 | 90.1 | 75.6 | |
| InternVL3.5-8BParadigm=Open-Source Multimodal Models2026.06 | 40.9 | 88.9 | 72.4 | |
| Qwen3-VL-8BParadigm=Open-Source Multimodal Models2026.06 | 40.1 | 86.8 | 69.5 | |
| LLaVA-Med-1.5Paradigm=Open-Source Multimodal Models2026.06 | 38.8 | 82.1 | 61.4 | |
| METransformerParadigm=Report Generation Baselines2026.06 | 34.1 | 70.3 | 47 | |
| R2Gen-MambaParadigm=Report Generation Baselines2026.06 | 32.9 | 68.4 | 45.8 | |
| MCLN-TransformerParadigm=Report Generation Baselines2026.06 | 31.6 | 66.2 | 43.2 |