Recurrent Structural Policy Gradient for Partially Observable Mean Field Games
Clarisse Wibault, Johannes Forkel, Sebastian Towers, Tiphaine Wibault, Juan Duque, George Whittle, Andreas Schaab, Yucheng Yang, Chiyuan Wang, Michael Osborne, Benjamin Moll, Jakob Foerster · 2026
Read Recurrent Structural Policy Gradient for Partially Observable Mean Field Games by Clarisse Wibault, Johannes Forkel, Sebastian Towers, Tiphaine Wibault, Juan Duque, George Whittle, Andreas Schaab, Yucheng Yang, Chiyuan Wang, Michael Osborne, Benjamin Moll, Jakob Foerster (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-20141.txt.
Description
Mean field games model interactions in large populations. Algorithmic progress has lagged. Model-free methods suffer high variance. Exact methods scale poorly. Hybrid structural methods cut variance while staying tractable, but until now they had not covered partial observability. A 2026 arXiv paper by Clarisse Wibault and eleven co-authors extends them to that setting.
Hybrid structural methods exploit low-dimensional individual state and action spaces plus known transition dynamics. From those, they compute the exact expected return given Monte Carlo rollouts of common noise. The paper proposes Recurrent Structural Policy Gradient, the first history-aware hybrid structural method for mean field games with partial information. It converges an order of magnitude faster than model-free reinforcement learning methods, and it learns history-aware behaviour where hybrid structural methods do not. The authors introduce MFAX, a JAX-based framework for mean field games supporting analytic and sample mean-field updates.
Catalog record
| CREATOR | Clarisse Wibault, Johannes Forkel, Sebastian Towers, Tiphaine Wibault, Juan Duque, George Whittle, Andreas Schaab, Yucheng Yang, Chiyuan Wang, Michael Osborne, Benjamin Moll, Jakob Foerster |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 73 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-20141