Pretrained world model
A world model, one step ahead.
A pretrained TD-MPC2 model imagines a walker’s future in latent space. Compare its reward forecasts with real MuJoCo simulation under exactly the same actions.
Experiment
Remote GPU · recordedRecorded replay · No GPU needed to view. New runs require a remote GPU.
No recording
No recording has been published for this experiment yet.
Replay an operator-recorded GPU run. Public compute launch is not available.