Multimodal Deep Learning on VALSE actant swap
Metric: pairwise accuracy (higher is better)
Metric: pairwise accuracy (higher is better)
| # | Model↕ | pairwise accuracy▼ | Augmentations | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | GPT2 | 76.9 | No | VALSE: A Task-Independent Benchmark for Vision a... | 2021-12-14 | Code |
| 2 | GPT1 | 72.2 | No | VALSE: A Task-Independent Benchmark for Vision a... | 2021-12-14 | Code |
| 3 | CLIP | 68.6 | No | VALSE: A Task-Independent Benchmark for Vision a... | 2021-12-14 | Code |
| 4 | ViLBERT | 68.3 | No | VALSE: A Task-Independent Benchmark for Vision a... | 2021-12-14 | Code |
| 5 | ViLBERT 12-in-1 | 58.9 | No | VALSE: A Task-Independent Benchmark for Vision a... | 2021-12-14 | Code |
| 6 | LXMERT | 45.8 | No | VALSE: A Task-Independent Benchmark for Vision a... | 2021-12-14 | Code |
| 7 | VisualBERT | 44.4 | No | VALSE: A Task-Independent Benchmark for Vision a... | 2021-12-14 | Code |