Action Prediction on ProcTHOR single-object (OOD Compositional)
95AccuracyCDE (ViT-MAE)
Evaluation Results
| Method | Links | |
|---|---|---|
| CDE (ViT-MAE)Backbone=ViT, Pre-training=MAE2025.08 | 95 | |
| CDE (ViT-CLIP)Backbone=ViT, Pre-training=CLIP2025.08 | 91 | |
| CDE (ViT-DINO)Backbone=ViT, Pre-training=DINO2025.08 | 91 | |
| SMS-RBackbone=ResNet-182025.08 | 47 | |
| ICM-RBackbone=ResNet-182025.08 | 41 | |
| ICM-VBackbone=ViT-DINO2025.08 | 38 | |
| Vanilla-RBackbone=ResNet-182025.08 | 36 | |
| Vanilla-VBackbone=ViT-DINO2025.08 | 34 | |
| SMS-VBackbone=ViT-DINO2025.08 | 34 |