Decision Transformer - Model Free or Model Based?

Decision transformer is a model based model like RL algorithm. It might need the right kind of latent space to predict the future it's not ideal for you to just predict the future pixels in Atari. The ideas like Dolly do are getting at like you just like learn a prior on the latent space and then you decode it with the diffusion of sampler. Those are really cool. Yeah. So we've talked about the difference between the RL classical RL paradigm and the supervised learning paradigm and the decision transformer kind of combines those what about the axis of model free and model based? yeah.