| Paper | Code | R@1 | ModelName | ReleaseDate |
|---|---|---|---|---|
| Hyperbolic Gramian Volumes for Multimodal Alignment | 53.6 | HyperGRAM (extra data) | 2026-06-03 | |
| Perception Encoder: The best visual embeddings are not at the output of the network | ✓ Link | 49.9 | PE-Core G | 2025-04-17 |
| MoE-ViE: Mixture of Experts Vision Encoder for Efficient Image and Video Understanding | ✓ Link | 49.8 | MoE-ViE-L/16 | 2026-08-18 |
| LAION-BVD: A 10-Million-Hour Open Video Dataset for Multimodal Pre-training | 45.3 | ViCLIP L-14 (BVD-V-10M, WiSE-FT, zero-shot) | 2026-08-25 | |
| Incorporating Dense Knowledge Alignment into Unified Multimodal Representation Models | 43.7 | DeKR-7B (zero-shot) | 2025-06-11 |