Saved in:
Bibliographic Details
Main Authors: Li, Ziming, Wu, Xiaoming, Wang, Zehong, Li, Jiazheng, Tian, Yijun, Bi, Jinhe, Ma, Yunpu, Ye, Yanfang, Zhang, Chuxu
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.22384
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911717495144448
author Li, Ziming
Wu, Xiaoming
Wang, Zehong
Li, Jiazheng
Tian, Yijun
Bi, Jinhe
Ma, Yunpu
Ye, Yanfang
Zhang, Chuxu
author_facet Li, Ziming
Wu, Xiaoming
Wang, Zehong
Li, Jiazheng
Tian, Yijun
Bi, Jinhe
Ma, Yunpu
Ye, Yanfang
Zhang, Chuxu
contents Graphs provide a natural representation of relational structure that arises across diverse domains. Despite this ubiquity, graph structure is typically learned in a modality- and task-isolated manner, where graph representations are constructed within individual task contexts and discarded thereafter. As a result, structural regularities across modalities and tasks are repeatedly reconstructed rather than accumulated at the level of intermediate graph representations. This motivates a representation-learning question: how should graph structure be organized so that it can persist and accumulate across heterogeneous modalities and tasks? We adopt a representation-centric perspective in which graph structure is treated as a structural substrate that persists across learning contexts. To instantiate this perspective, we propose G-Substrate, a graph substrate framework that organizes learning around shared graph structures. G-Substrate comprises two complementary mechanisms: a unified structural schema that ensures compatibility among graph representations across heterogeneous modalities and tasks, and an interleaved role-based training strategy that exposes the same graph structure to multiple functional roles during learning. Experiments across multiple domains, modalities, and tasks show that G-Substrate outperforms task-isolated and naive multi-task learning methods. The codebase, model, and datasets are available at https://github.com/zmli6/G-Substrate.
format Preprint
id arxiv_https___arxiv_org_abs_2601_22384
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Graph is a Substrate Across Data Modalities
Li, Ziming
Wu, Xiaoming
Wang, Zehong
Li, Jiazheng
Tian, Yijun
Bi, Jinhe
Ma, Yunpu
Ye, Yanfang
Zhang, Chuxu
Machine Learning
Artificial Intelligence
Graphs provide a natural representation of relational structure that arises across diverse domains. Despite this ubiquity, graph structure is typically learned in a modality- and task-isolated manner, where graph representations are constructed within individual task contexts and discarded thereafter. As a result, structural regularities across modalities and tasks are repeatedly reconstructed rather than accumulated at the level of intermediate graph representations. This motivates a representation-learning question: how should graph structure be organized so that it can persist and accumulate across heterogeneous modalities and tasks? We adopt a representation-centric perspective in which graph structure is treated as a structural substrate that persists across learning contexts. To instantiate this perspective, we propose G-Substrate, a graph substrate framework that organizes learning around shared graph structures. G-Substrate comprises two complementary mechanisms: a unified structural schema that ensures compatibility among graph representations across heterogeneous modalities and tasks, and an interleaved role-based training strategy that exposes the same graph structure to multiple functional roles during learning. Experiments across multiple domains, modalities, and tasks show that G-Substrate outperforms task-isolated and naive multi-task learning methods. The codebase, model, and datasets are available at https://github.com/zmli6/G-Substrate.
title Graph is a Substrate Across Data Modalities
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2601.22384