Relational Reasoning on Sort-of-CLEVR (test)
80.03Relational AccuracyAiT-Base
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AiT-BaseModel architecture=Transformer based, Model scale=Base2023.09 | 80.03 | 99.98 | |
| AiT-MediumModel architecture=Transformer based, Model scale=Medium2023.09 | 78.14 | 99.75 | |
| AiT-SmallModel architecture=Transformer based, Model scale=Small2023.09 | 76.82 | 99.85 | |
| CoordinationModel architecture=Transformer based2023.09 | 73.43 | 96.31 | |
| ViT-BaseModel architecture=Vision Transformer, Model scale=Base2023.09 | 63.35 | 99.73 | |
| CNN+LSTMModel architecture=Non-Transformer based2023.09 | 60.08 | 99.47 | |
| ViT-MediumModel architecture=Vision Transformer, Model scale=Medium2023.09 | 54.71 | 99.7 | |
| ViT-SmallModel architecture=Vision Transformer, Model scale=Small2023.09 | 51.75 | 98.8 | |
| Set TransformerModel architecture=Transformer based2023.09 | 47.63 | 57.65 | |
| Dense-SmallModel architecture=Non-Transformer based (Fully-connected), Model scale=Small2023.09 | 47.28 | 57.68 | |
| Dense-BaseModel architecture=Non-Transformer based (Fully-connected), Model scale=Base2023.09 | 46.93 | 57.71 |