Regression on Moonstone Cross-modal (Downstream)
0.987R2MG-MAE
Evaluation results
| Method | Links | |
|---|---|---|
| MG-MAEPre-training=Lunar, Evaluation Protocol=Finetune2026.07 | 0.987 | |
| MG-MAEPre-training=Lunar, Evaluation Protocol=Linear2026.07 | 0.971 | |
| Vanilla MAEPre-training=Lunar, Details=Per-channel tokenizers without modality grouping2026.07 | 0.956 | |
| ScratchPre-training=None2026.07 | 0.941 | |
| ImageNet ViT-BPre-training=ImageNet, Evaluation Protocol=Linear2026.07 | 0.812 |