{"ID":2879244,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2508.16027","arxiv_id":"2508.16027","title":"Optimal Dynamic Regret by Transformers for Non-Stationary Reinforcement Learning","abstract":"Transformers have demonstrated exceptional performance across a wide range of domains. While their ability to perform reinforcement learning in-context has been established both theoretically and empirically, their behavior in non-stationary environments remains less understood. In this study, we address this gap by showing that transformers can achieve nearly optimal dynamic regret bounds in non-stationary settings. We prove that transformers are capable of approximating strategies used to handle non-stationary environments and can learn the approximator in the in-context learning setup. Our experiments further show that transformers can match or even outperform existing expert algorithms in such environments.","short_abstract":"Transformers have demonstrated exceptional performance across a wide range of domains. While their ability to perform reinforcement learning in-context has been established both theoretically and empirically, their behavior in non-stationary environments remains less understood. In this study, we address this gap by sh...","url_abs":"https://arxiv.org/abs/2508.16027","url_pdf":"https://arxiv.org/pdf/2508.16027v2","authors":"[\"Baiyuan Chen\",\"Shinji Ito\",\"Masaaki Imaizumi\"]","published":"2025-08-22T01:07:32Z","proceeding":"stat.ML","tasks":"[\"stat.ML\",\"cs.LG\"]","methods":"[\"Reinforcement Learning\",\"Transformer\"]","has_code":false}