Transformers are Sample-Efficient World Models
This addresses the sample inefficiency issue for reinforcement learning agents, enabling more practical applications in real-world scenarios, though it is an incremental improvement building on existing model-based approaches.
The paper tackles the sample inefficiency problem in deep reinforcement learning by introducing IRIS, a model-based agent that uses a discrete autoencoder and an autoregressive Transformer as a world model. It achieves a mean human normalized score of 1.046 on the Atari 100k benchmark with only two hours of gameplay, outperforming humans on 10 out of 26 games and setting a new state of the art for methods without lookahead search.
Deep reinforcement learning agents are notoriously sample inefficient, which considerably limits their application to real-world problems. Recently, many model-based methods have been designed to address this issue, with learning in the imagination of a world model being one of the most prominent approaches. However, while virtually unlimited interaction with a simulated environment sounds appealing, the world model has to be accurate over extended periods of time. Motivated by the success of Transformers in sequence modeling tasks, we introduce IRIS, a data-efficient agent that learns in a world model composed of a discrete autoencoder and an autoregressive Transformer. With the equivalent of only two hours of gameplay in the Atari 100k benchmark, IRIS achieves a mean human normalized score of 1.046, and outperforms humans on 10 out of 26 games, setting a new state of the art for methods without lookahead search. To foster future research on Transformers and world models for sample-efficient reinforcement learning, we release our code and models at https://github.com/eloialonso/iris.