Predicting sticker colors directly
We first replace video prediction with prediction of the 12 visible sticker colors. A small Transformer receives the initial visible configuration and the same kind of action-prompt features used by the video models. Partial observation remains: a sticker can disappear and later return, so the model still has to account for intervening moves.
Video space
Wan VAE
Vector space
encoding
These state-only models predict sequences of 20 actions with high accuracy at much lower compute. Their success shows that the cube’s transitions can be learned efficiently in this representation. The additional difficulty appears when those transitions must be represented and generated through video.
| State model | Training PF-days | Final-state accuracy | Full-trajectory accuracy |
|---|---|---|---|
| AR · 2.8M | ≈ 0.008 | 94.01% | 92.81% |
| Bidirectional · 2.8M | ≈ 0.0004 | 99.86% | 99.85% |
Final-state accuracy requires all visible sticker colors to be correct after move 20. Full-trajectory accuracy requires every predicted state to be correct. These are separate state-only experiments.

