Motion localization on MeViS
68SL ScoreIMAP
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| IMAPBackbone=CogVideoX-5B2026.03 | 68 | 48 | 69 | 61 | 64 | 62 | |
| IMAPBackbone=HunyuanVideo2026.03 | 60 | 41 | 62 | 50 | 62 | 55 | |
| IMAPBackbone=Wan2.1-14B2026.03 | 58 | 41 | 60 | 47 | 49 | 51 | |
| ConceptAttentionBackbone=CogVideoX-5B2026.03 | 50 | 32 | 51 | 47 | 47 | 45 | |
| IMAPBackbone=Wan2.1-1.3B2026.03 | 50 | 33 | 52 | 40 | 40 | 43 | |
| Cross AttentionBackbone=Wan2.1-14B2026.03 | 50 | 32 | 51 | 42 | 41 | 43 | |
| IMAPBackbone=CogVideoX-2B2026.03 | 49 | 62 | 56 | 48 | 55 | 54 | |
| Cross AttentionBackbone=Wan2.1-1.3B2026.03 | 49 | 32 | 45 | 43 | 45 | 43 | |
| ConceptAttentionBackbone=HunyuanVideo2026.03 | 42 | 26 | 44 | 35 | 34 | 36 | |
| ConceptAttentionBackbone=CogVideoX-2B2026.03 | 42 | 26 | 44 | 35 | 34 | 36 | |
| Cross AttentionBackbone=CogVideoX-5B2026.03 | 41 | 27 | 43 | 34 | 33 | 36 | |
| Cross AttentionBackbone=HunyuanVideo2026.03 | 39 | 25 | 41 | 36 | 34 | 35 | |
| DAAMBackbone=VideoCrafter22026.03 | 36 | 17 | 38 | 32 | 35 | 32 | |
| ViCLIPBackbone=ViT-H2026.03 | 33 | 17 | 35 | 29 | 28 | 28 | |
| Cross AttentionBackbone=CogVideoX-2B2026.03 | 29 | 56 | 35 | 23 | 28 | 34 |