Abstractive Summarization on YouCook2 (test)
33ROUGE-1ClipSum
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ClipSumVisual Features=CLIP ViT-B/32 (512-dim)2026.05 | 33 | 6.4 | 25.3 | 10.8 | 22.1 | |
| VG-BARTVisual Features=ResNet (2048-dim)2026.05 | 30.8 | 5.3 | 24.2 | 9.7 | 20.6 | |
| BART + ResNet-152Visual Features=ImageNet (2048-dim)2026.05 | 30.5 | 5.1 | 23.9 | 9.5 | 20.3 | |
| MFFG-TransformerVisual Features=ResNet (2048-dim)2026.05 | 30.1 | 4.8 | 23.6 | 9.2 | 19.8 | |
| HA-TransformerVisual Features=ResNet (2048-dim)2026.05 | 29.4 | 4.3 | 23.1 | 8.9 | 19.2 | |
| BART + Random VisualVisual Features=Random (512-dim)2026.05 | 29.1 | 4.2 | 22.8 | 8.7 | 19.1 | |
| BARTVisual Features=–2026.05 | 28.3 | 3.8 | 22.1 | 8.2 | 18.5 | |
| T5Visual Features=–2026.05 | 27.8 | 3.5 | 21.6 | 7.9 | 17.9 |