Object Dynamics Prediction on YouTube-VIS
66.6FG-ARIDSSA (identity)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DSSA (identity)Representation=identity state (e^k_t), Dynamics Model=SlotFormer [29], Representations=Frozen2026.06 | 66.6 | 52.7 | 51.4 | |
| DSSA (local)Representation=local state (l^k_t), Dynamics Model=SlotFormer [29], Representations=Frozen2026.06 | 66.5 | 52.6 | 51.2 | |
| RandSF.QtsimDynamics Model=SlotFormer [29], Representations=Frozen2026.06 | 46.6 | 43.7 | 43.1 | |
| SlotContrastDynamics Model=SlotFormer [29], Representations=Frozen2026.06 | 37.9 | 33.2 | 33.1 | |
| SlotCurriEvaluation Framework=SlotFormer (SF)2026.03 | 31.4 | 30.1 | — | |
| SContrastEvaluation Framework=SlotFormer (SF)2026.03 | 29.2 | 29.6 | — | |
| ReconEvaluation Framework=SlotFormer (SF)2026.03 | 27.4 | 28.9 | — |