English

A Survey on Transformers in Reinforcement Learning

Machine Learning 2023-09-22 v3 Artificial Intelligence

Abstract

Transformer has been considered the dominating neural architecture in NLP and CV, mostly under supervised settings. Recently, a similar surge of using Transformers has appeared in the domain of reinforcement learning (RL), but it is faced with unique design choices and challenges brought by the nature of RL. However, the evolution of Transformers in RL has not yet been well unraveled. In this paper, we seek to systematically review motivations and progress on using Transformers in RL, provide a taxonomy on existing works, discuss each sub-field, and summarize future prospects.

Keywords

Cite

@article{arxiv.2301.03044,
  title  = {A Survey on Transformers in Reinforcement Learning},
  author = {Wenzhe Li and Hao Luo and Zichuan Lin and Chongjie Zhang and Zongqing Lu and Deheng Ye},
  journal= {arXiv preprint arXiv:2301.03044},
  year   = {2023}
}

Comments

Accepted by TMLR

R2 v1 2026-06-28T08:06:41.119Z