Stabilizing Autoregressive Forecasts in Chaotic Systems via Multi-rate Latent Recurrence

Mrigank Dhingra, Omer San · 2026

Read Stabilizing Autoregressive Forecasts in Chaotic Systems via Multi-rate Latent Recurrence by Mrigank Dhingra, Omer San (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-2601-14487.txt.

Description

Mrigank Dhingra and Omer San work on forecasting chaotic systems in this 2026 paper. Long-horizon autoregressive forecasting of chaotic dynamics is hard for two reasons. Small one-step errors amplify fast. Rollouts also drift off distribution, so predictions collapse into physically inconsistent states with wrong large-scale statistics.

The team introduces MSR-HINE, a hierarchical implicit forecaster. It augments multiscale latent priors with multi-rate recurrent modules. Those modules run at distinct temporal scales, fast and slow. The design aims to stabilize long rollouts. The point is trust. A long run that stays true to the real system is worth more than a short one that drifts. Readers get a forecaster built to keep the broad statistics of a chaotic system intact over long runs.

Catalog record

CREATORMrigank Dhingra, Omer San
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE79 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-14487

Browse the full library · About the Tower of Babel