Saved in:
Bibliographic Details
Main Authors: Lnu, Aakriti, Li, Zhe, Liang, Dandan, Huang, Chao, Li, Rui, Yang, Haibo
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.10940
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918427058241536
author Lnu, Aakriti
Li, Zhe
Liang, Dandan
Huang, Chao
Li, Rui
Yang, Haibo
author_facet Lnu, Aakriti
Li, Zhe
Liang, Dandan
Huang, Chao
Li, Rui
Yang, Haibo
contents Split learning (SL) enables collaborative training of large language models (LLMs) between resource-constrained edge devices and compute-rich servers by partitioning model computation across the network boundary. However, existing SL systems predominantly rely on first-order (FO) optimization, which requires clients to store intermediate quantities such as activations for backpropagation. This results in substantial memory overhead, largely negating benefits of model partitioning. In contrast, zeroth-order (ZO) optimization eliminates backpropagation and significantly reduces memory usage, but often suffers from slow convergence and degraded performance. In this work, we propose HOSL, a novel Hybrid-Order Split Learning framework that addresses this fundamental trade-off between memory efficiency and optimization effectiveness by strategically integrating ZO optimization on the client side with FO optimization on the server side. By employing memory-efficient ZO gradient estimation at the client, HOSL eliminates backpropagation and activation storage, reducing client memory consumption. Meanwhile, server-side FO optimization ensures fast convergence and competitive performance. Theoretically, we show that HOSL achieves an $\mathcal{O}(\sqrt{d_c/TQ})$ rate, which depends on client-side model dimension $d_c$ rather than the full model dimension $d$, demonstrating that convergence improves as more computation is offloaded to the server. Extensive experiments on OPT models (125M and 1.3B parameters) across 6 tasks demonstrate that HOSL reduces client GPU memory by up to 3.7$\times$ compared to the FO method while achieving accuracy within 0.20%-4.23% of this baseline. Furthermore, HOSL outperforms the ZO baseline by up to 15.55%, validating the effectiveness of our hybrid strategy for memory-efficient training on edge devices.
format Preprint
id arxiv_https___arxiv_org_abs_2601_10940
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle HOSL: Hybrid-Order Split Learning for Memory-Constrained Edge Training
Lnu, Aakriti
Li, Zhe
Liang, Dandan
Huang, Chao
Li, Rui
Yang, Haibo
Machine Learning
Split learning (SL) enables collaborative training of large language models (LLMs) between resource-constrained edge devices and compute-rich servers by partitioning model computation across the network boundary. However, existing SL systems predominantly rely on first-order (FO) optimization, which requires clients to store intermediate quantities such as activations for backpropagation. This results in substantial memory overhead, largely negating benefits of model partitioning. In contrast, zeroth-order (ZO) optimization eliminates backpropagation and significantly reduces memory usage, but often suffers from slow convergence and degraded performance. In this work, we propose HOSL, a novel Hybrid-Order Split Learning framework that addresses this fundamental trade-off between memory efficiency and optimization effectiveness by strategically integrating ZO optimization on the client side with FO optimization on the server side. By employing memory-efficient ZO gradient estimation at the client, HOSL eliminates backpropagation and activation storage, reducing client memory consumption. Meanwhile, server-side FO optimization ensures fast convergence and competitive performance. Theoretically, we show that HOSL achieves an $\mathcal{O}(\sqrt{d_c/TQ})$ rate, which depends on client-side model dimension $d_c$ rather than the full model dimension $d$, demonstrating that convergence improves as more computation is offloaded to the server. Extensive experiments on OPT models (125M and 1.3B parameters) across 6 tasks demonstrate that HOSL reduces client GPU memory by up to 3.7$\times$ compared to the FO method while achieving accuracy within 0.20%-4.23% of this baseline. Furthermore, HOSL outperforms the ZO baseline by up to 15.55%, validating the effectiveness of our hybrid strategy for memory-efficient training on edge devices.
title HOSL: Hybrid-Order Split Learning for Memory-Constrained Edge Training
topic Machine Learning
url https://arxiv.org/abs/2601.10940