No Epoch Like the Present: Robust Climate Emulation Requires Out-of-Distribution Generalisation

Bradley Stanley-Clamp, Anson Lei, Hannah M. Christensen, Ingmar Posner · 2026

Read No Epoch Like the Present: Robust Climate Emulation Requires Out-of-Distribution Generalisation by Bradley Stanley-Clamp, Anson Lei, Hannah M. Christensen, Ingmar Posner (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-2605-22248.txt.

Description

Stanley-Clamp and colleagues argue that climate emulation is at heart an out-of-distribution task, since a warming world shifts atmospheric patterns beyond training data. They show seasonal change works as a real-world proxy for those long-term shifts. Current hybrid machine learning emulators degrade sharply under such seasonal tests. The authors point to compositional generalisation, recombining observed parts in new ways, as the principled route forward. Physically motivated decompositions already lift out-of-distribution skill at modest cost.

Catalog record

CREATORBradley Stanley-Clamp, Anson Lei, Hannah M. Christensen, Ingmar Posner
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE89130
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.22248

Browse the full library · About the Tower of Babel