HOPE
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Continual Language Modeling | HOPE 8-domain, clean-regime 256M (3 seeds) | Avg PPL35.8 | 19 | |
| Monocular Depth Estimation | HOPE | Delta-1 Accuracy61.36 | 16 | |
| Continual Learning | HOPE 8-domain (test) | Forgetting9.1 | 9 | |
| Semantic Segmentation | HOPE | mIoU6.9 | 8 | |
| Continual Language Modeling | HOPE 8-domain, high-overlap non-adaptive stream 256M | Forgetting9.4 | 6 | |
| Object Detection | HOPE BOP v2 (test) | AP41.1 | 5 | |
| Continual Language Modeling | HOPE 16-domain stream 256M (long continual sequence) | Forgetting10.2 | 4 | |
| 6DoF Pose Estimation | HOPE BOP v2 (test) | AP33.5 | 4 | |
| Camera Pose Estimation | HOPE v2 | R Error @50.095 | 3 |
Showing 9 of 9 rows