DINO-WM（ICML 2025）

DINO-WM models the world dynamics on compact embeddings of the world, rather than the raw observations themselves.

This allows DINO-WM to achieve observational goals through action sequence optimization, facilitating task-agnostic planning by treating goal features as prediction targets.

Given these visual embeddings and actions, DINO-WM uses the ViT architecture to predict future embeddings.

DINO-WM leverages spatial patch features pre-trained with DINOv2, enabling it to learn from offline behavioral trajectories by predicting future patch features.

We have uploaded our trained world model checkpoints for PointMaze, PushT, and Wall here under checkpoints.