Table 4 Semantic answer similarity validation.
From: Multi-modal transformer architecture for medical image analysis and automated report generation
Model | Skip thought CS | Vector extrema | Greedy matching |
|---|---|---|---|
BEiTGPT2 | 0.9829 | 0.9871 | 0.9878 |
DEiTGPT2 | 0.9761 | 0.98 | 0.9813 |
ViTGPT2 | 0.9811 | 0.9836 | 0.9882 |