Abstract Visual Reasoning on SVRT reformulated four-choice (test)
99.4AccuracyPR-A2CL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PR-A2CLTraining Sample Size=10000, Backbone=PARM2026.03 | 99.4 | 86.7 | |
| DBCRTraining Sample Size=10000, Backbone=DBCR2026.03 | 98.8 | 81 | |
| R3PCLTraining Sample Size=10000, Backbone=R3PCL2026.03 | 98.5 | 79.3 | |
| PR-A2CLTraining Sample Size=1000, Backbone=PARM2026.03 | 98.2 | 86.7 | |
| PR-A2CLTraining Sample Size=500, Backbone=PARM2026.03 | 96.9 | 86.7 | |
| IN-ViT-SmallTraining Sample Size=10000, Backbone=ViT-Small, Pre-training=ImageNet2026.03 | 92.4 | 27.5 | |
| CLIP-ViT-BaseTraining Sample Size=1000, Backbone=ViT-Base, Pre-training=CLIP2026.03 | 92.1 | 57.9 | |
| PR-A2CLTraining Sample Size=200, Backbone=PARM2026.03 | 92.1 | 86.7 | |
| CLIP-ResNet-50Training Sample Size=1000, Backbone=ResNet-50, Pre-training=CLIP2026.03 | 89.3 | 55.7 | |
| PR-A2CLTraining Sample Size=100, Backbone=PARM2026.03 | 85.9 | 86.7 | |
| SSL-ResNet-50Training Sample Size=1000, Backbone=ResNet-50, Pre-training=MoCo-v32026.03 | 83.5 | 69.2 | |
| IN-ResNet-50Training Sample Size=1000, Backbone=ResNet-50, Pre-training=ImageNet2026.03 | 83 | 53.8 | |
| PR-A2CLTraining Sample Size=50, Backbone=PARM2026.03 | 78.5 | 86.7 | |
| WReNTraining Sample Size=10000, Backbone=WReN2026.03 | 70.3 | 30.8 | |
| PR-A2CLTraining Sample Size=20, Backbone=PARM2026.03 | 68.5 | 86.7 | |
| SCL-ResNet-18Training Sample Size=1000, Backbone=ResNet-182026.03 | 68.2 | 38.2 | |
| SSL-ViT-SmallTraining Sample Size=1000, Backbone=ViT-Small, Pre-training=MoCo-v32026.03 | 67.1 | 53.3 | |
| DBCRTraining Sample Size=20, Backbone=DBCR2026.03 | 65.2 | 81 | |
| R3PCLTraining Sample Size=20, Backbone=R3PCL2026.03 | 64.8 | 79.3 | |
| SCLTraining Sample Size=1000, Backbone=SCL2026.03 | 55 | 31.7 | |
| ResNet-50Training Sample Size=1000, Backbone=ResNet-50, Pre-training=None2026.03 | 54 | 37.1 | |
| PredRNetTraining Sample Size=1000, Backbone=PredRNet2026.03 | 51.3 | 38.3 | |
| SCARTraining Sample Size=10000, Backbone=SCAR2026.03 | 50.3 | 28.5 | |
| ViT-SmallTraining Sample Size=10000, Backbone=ViT-Small, Pre-training=None2026.03 | 40 | 26.4 | |
| WReNTraining Sample Size=20, Backbone=WReN2026.03 | 26 | 30.8 | |
| SCARTraining Sample Size=20, Backbone=SCAR2026.03 | 25.8 | 28.5 | |
| IN-ViT-SmallTraining Sample Size=20, Backbone=ViT-Small, Pre-training=ImageNet2026.03 | 25.6 | 27.5 | |
| ViT-SmallTraining Sample Size=20, Backbone=ViT-Small, Pre-training=None2026.03 | 24.7 | 26.4 |