arXiv 2507.12549
The Serial Scaling Hypothesis
By Yuxi Liu, Konpat Preechakul, et al.
Published 2025-07-16
Citation lineage
Review the prior work and downstream research connected to this paper.
While machine learning has advanced through massive parallelization, we identify a critical blind spot: some problems are fundamentally sequential. These "inherently serial" problems-from mathematical reasoning to physical simulations to sequential decision-making-require sequentially dependent computational steps that cannot be efficiently parallelized. We formalize this distinction in complexity theory, and demons…