2024/06/27 by Vincent Micheli, Eloi Alonso, Micheli, Vincent +3 · 5 citations
Computer Science · #Advanced Database Systems and Queries #Artificial Intelligence (cs.AI) #Computer Vision and Pattern Recognition (cs.CV) #Data Management and Algorithms #FOS: Computer and information sciences #Graph Theory and Algorithms #Machine Learning (cs.LG)
paper · pdf · doi:10.48550/arxiv.2406.19320
openalex publication_date 2024/06/27 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
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 Δ-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, Δ-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.