Unraveling the Hidden Dynamical Structure in Recurrent Neural Policies

Jin Li, Yue Wu, Mengsha Huang, Yuhao Sun, Hao He, Xianyuan Zhan · 2026

Read Unraveling the Hidden Dynamical Structure in Recurrent Neural Policies by Jin Li, Yue Wu, Mengsha Huang, Yuhao Sun, Hao He, Xianyuan Zhan (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-2602-01196.txt.

Description

Jin Li and co-authors find hidden loop shapes in the inner states of recurrent control policies. Such policies hold memory and adapt fast, beating plain policies, yet why they travel so well stayed murky. Across train plans, net forms, and jobs, the team finds steady cycles that pop up as policies meet their worlds, much like limit cycles in dynamical lore. Those cycles steady both inner memory and job-relevant world states, mute nuisance shifts, and their shapes encode act ties that ease skill shifts in changing worlds.

Catalog record

CREATORJin Li, Yue Wu, Mengsha Huang, Yuhao Sun, Hao He, Xianyuan Zhan
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE88438
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.01196

Browse the full library · About the Tower of Babel