Panel Captioning on MedICaT
7.83BLEU-4FigEx2-8B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| FigEx2-8BProtocol=Fine-tuned2026.01 | 7.83 | 19.74 | 13.89 | 86.49 | |
| Qwen3-VL-8BProtocol=Fine-tuned2026.01 | 7.37 | 18.13 | 12.78 | 86.25 | |
| Llama3.2-11BProtocol=Fine-tuned2026.01 | 4.8 | 16.38 | 11.39 | 85.62 | |
| LLaVA-NeXTProtocol=Fine-tuned2026.01 | 4.06 | 14.27 | 9.28 | 85.14 | |
| Llama3.2-90BProtocol=Pre-trained2026.01 | 3 | 11.73 | 7.97 | 85.68 | |
| gpt-5-miniProtocol=Pre-trained2026.01 | 2.92 | 14.44 | 9.38 | 85.44 | |
| gpt-5.2Protocol=Pre-trained2026.01 | 2.43 | 13.25 | 7.44 | 85.54 | |
| LLaVA-MedProtocol=Fine-tuned2026.01 | 2.26 | 10.58 | 6.26 | 85 | |
| Qwen2.5-VL-72BProtocol=Pre-trained2026.01 | 2.08 | 15.91 | 9.95 | 83.67 | |
| gpt-5-nanoProtocol=Pre-trained2026.01 | 2.01 | 10.76 | 5.91 | 85.23 |