Multimodal Abstractive Summarization on MM-Sum Serbian Latin, low-resource
21.89ROUGE-1SOV-MAS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SOV-MASTraining Setting=Multilingual Training, Visual Features=ViT2022.12 | 21.89 | 6.81 | 18.32 | |
| mT5Training Setting=Multilingual Training, Visual Features=ViT2022.12 | 21.28 | 6.04 | 17.41 | |
| VG-mT5Training Setting=Multilingual Training, Visual Features=ViT2022.12 | 20.54 | 5.8 | 17.2 | |
| SOV-MASTraining Setting=Monolingual Training, Visual Features=ViT2022.12 | 18.58 | 4.88 | 15.42 | |
| VG-mT5Training Setting=Monolingual Training, Visual Features=ViT2022.12 | 18.55 | 4.69 | 15.53 | |
| mT5Training Setting=Monolingual Training, Visual Features=ViT2022.12 | 18.5 | 4.4 | 15.11 |