Multi-entity Video Segmentation on MeViS
51.3Combined J&FGLUS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GLUSBackbone=Chat-UniVi-7B2026.06 | 51.3 | 48.5 | 54.2 | |
| VRS-HQBackbone=Chat-UniVi-7B2026.06 | 50.6 | 47.6 | 53.7 | |
| FeVOS-R1Backbone=InternVL2.5-4B2026.06 | 49.5 | 46.2 | 52.7 | |
| Sa2VA*Backbone=InternVL2.5-4B2026.06 | 46.5 | 43.4 | 49.7 | |
| Sa2VABackbone=InternVL2.5-4B2026.06 | 46.4 | — | — | |
| VideoGLaMMBackbone=Phi3-Mini-3.8B2026.06 | 45.2 | 42.1 | 48.2 | |
| VideoLISABackbone=LLaVA-Phi-3-V-3.8B2026.06 | 44.4 | 41.3 | 47.6 | |
| VISABackbone=Chat-UniVi-7B2026.06 | 43.5 | 40.7 | 46.3 | |
| TrackGPTBackbone=LLaVA-7B2026.06 | 40.1 | 37.6 | 42.6 | |
| LMPMBackbone=Swin-T2026.06 | 37.2 | 34.2 | 40.2 | |
| LISABackbone=LLaVA-7B2026.06 | 37.2 | 35.1 | 39.4 | |
| ReferFormerBackbone=Video-Swin-B2026.06 | 31 | 29.8 | 32.2 |