Sliding Window Recurrences for Sequence Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Secrieru, Dragos, Brixi, Garyk, Bengio, Yoshua, Suzuki, Taiji, Poli, Michael, Massaroli, Stefano
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918249684271104
author Secrieru, Dragos
Brixi, Garyk
Bengio, Yoshua
Suzuki, Taiji
Poli, Michael
Massaroli, Stefano
author_facet Secrieru, Dragos
Brixi, Garyk
Bengio, Yoshua
Suzuki, Taiji
Poli, Michael
Massaroli, Stefano
contents Multi-hybrid architectures are poised to take over language modeling due to better quality and performance. We introduce a hierarchical decomposition framework for linear recurrences that allows us to develop algorithms aligned with GPU memory hierarchies, yielding Sliding Window Recurrences. We focus specifically on truncating recurrences to hardware-aligned windows which are naturally jagged, limiting costly inter-warp communication. Using SWR, we develop Phalanx layers that serve as drop-in replacements for windowed attention or linear recurrences. In 1B parameter multi-hybrid models, Phalanx achieves over 10-40% speedup across 4K to 32K context length over optimized Transformers while matching perplexity.
format Preprint
id arxiv_https___arxiv_org_abs_2512_13921
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Sliding Window Recurrences for Sequence Models
Secrieru, Dragos
Brixi, Garyk
Bengio, Yoshua
Suzuki, Taiji
Poli, Michael
Massaroli, Stefano
Machine Learning
Multi-hybrid architectures are poised to take over language modeling due to better quality and performance. We introduce a hierarchical decomposition framework for linear recurrences that allows us to develop algorithms aligned with GPU memory hierarchies, yielding Sliding Window Recurrences. We focus specifically on truncating recurrences to hardware-aligned windows which are naturally jagged, limiting costly inter-warp communication. Using SWR, we develop Phalanx layers that serve as drop-in replacements for windowed attention or linear recurrences. In 1B parameter multi-hybrid models, Phalanx achieves over 10-40% speedup across 4K to 32K context length over optimized Transformers while matching perplexity.
title Sliding Window Recurrences for Sequence Models
topic Machine Learning
url https://arxiv.org/abs/2512.13921