Efficient World Models with Context-Aware Tokenization

TL;DR

This paper introduces $\Delta$-IRIS, a novel world model architecture that uses context-aware tokenization to improve efficiency and performance in deep RL, achieving state-of-the-art results with faster training.

Contribution

The paper presents a new agent with a world model combining a discrete autoencoder for delta encoding and an autoregressive transformer, significantly reducing training time and improving performance.

Findings

Sets new state-of-the-art in Crafter benchmark.

Faster training compared to previous attention-based models.

Effective environment simulation with fewer tokens.

Abstract

Scaling up deep Reinforcement Learning (RL) methods presents a significant challenge. Following developments in generative modelling, model-based RL positions itself as a strong contender. Recent advances in sequence modelling have led to effective transformer-based world models, albeit at the price of heavy computations due to the long sequences of tokens required to accurately simulate environments. In this work, we propose $\Delta$-IRIS, a new agent with a world model architecture composed of a discrete autoencoder that encodes stochastic deltas between time steps and an autoregressive transformer that predicts future deltas by summarizing the current state of the world with continuous tokens. In the Crafter benchmark, $\Delta$-IRIS sets a new state of the art at multiple frame budgets, while being an order of magnitude faster to train than previous attention-based approaches. We release our code and models at https://github.com/vmicheli/delta-iris.