| DINOv3 | ✓ Link | 83.3 | DINOv3 + label propagation | 2025-08-13 |
| TimeBridge: Self-Supervised Video Representation Learning via Start-End Joint Embedding and In-Between Frame Prediction | | 73.5 | TimeBridge (ViT-S/8, Kinetics-400, 400 epochs) | 2026-06-03 |
| Emerging Properties in Self-Supervised Vision Transformers | ✓ Link | 71.4 | DINO (ViT-B/8) | 2021-04-29 |
| Vision Pretraining for Dense Spatial Perception | ✓ Link | 70.0 | LingBot-Vision ViT-g + label propagation | 2026-07-06 |
| VideoMAC: Video Masked Autoencoders Meet ConvNets | ✓ Link | 68.4 | VideoMAC (CNXv2-S) (extra data) | 2024-02-29 |
| Recurrent Video Masked Autoencoders | | 65.6 | RVM-H | 2026-04-21 |
| Patch-level Representation Learning for Self-supervised Vision Transformers | | 62.7 | DINO + SelfPatch (ViT-S/16) | 2022-06-16 |
| Perception Encoder: The best visual embeddings are not at the output of the network | ✓ Link | 61.5 | PE-Spatial G | 2025-04-17 |
| Connecting Joint-Embedding Predictive Architecture with Contrastive Self-supervised Learning | | 58.3 | C-JEPA (ViT-L/16, frozen-representation VOS) | 2024-10-25 |
| Learning Correspondence from the Cycle-Consistency of Time | ✓ Link | 48.2 | TimeCycle (ResNet-50, VLOG self-supervised, 480px) (extra data) | 2019-03-18 |