Continuity Laws for Sequential Models

Annan Yu, Dongwei Lyu, N. Benjamin Erichson · 2026

Read Continuity Laws for Sequential Models by Annan Yu, Dongwei Lyu, N. Benjamin Erichson (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-2605-08539.txt.

Description

Annan Yu, Dongwei Lyu, and N. Benjamin Erichson ask whether sequence models that claim roots in continuous time actually behave smoothly as time steps shrink. A model counts as continuous when its outputs settle toward one smooth path under finer time sampling. They also build a score for how continuous a dataset's own time structure is.

One state space model, S4, behaves steadily, while Mamba's core can react sharply to input size and selective dynamics. Across test sets, matching model continuity to task continuity tracks better results. Finer sampling tricks can then boost both speed and quality. Sequence modeling gains a clear law linking smooth time behavior to real gains.

Catalog record

CREATORAnnan Yu, Dongwei Lyu, N. Benjamin Erichson
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE79 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.08539

Browse the full library · About the Tower of Babel