Global-Lens Transformers: Adaptive Token Mixing for Dynamic Link Prediction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zou, Tao, Wu, Chengfeng, Liao, Tianxi, Ye, Junchen, Du, Bowen
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912712398733312
author Zou, Tao
Wu, Chengfeng
Liao, Tianxi
Ye, Junchen
Du, Bowen
author_facet Zou, Tao
Wu, Chengfeng
Liao, Tianxi
Ye, Junchen
Du, Bowen
contents Dynamic graph learning plays a pivotal role in modeling evolving relationships over time, especially for temporal link prediction tasks in domains such as traffic systems, social networks, and recommendation platforms. While Transformer-based models have demonstrated strong performance by capturing long-range temporal dependencies, their reliance on self-attention results in quadratic complexity with respect to sequence length, limiting scalability on high-frequency or large-scale graphs. In this work, we revisit the necessity of self-attention in dynamic graph modeling. Inspired by recent findings that attribute the success of Transformers more to their architectural design than attention itself, we propose GLFormer, a novel attention-free Transformer-style framework for dynamic graphs. GLFormer introduces an adaptive token mixer that performs context-aware local aggregation based on interaction order and time intervals. To capture long-term dependencies, we further design a hierarchical aggregation module that expands the temporal receptive field by stacking local token mixers across layers. Experiments on six widely-used dynamic graph benchmarks show that GLFormer achieves SOTA performance, which reveals that attention-free architectures can match or surpass Transformer baselines in dynamic graph settings with significantly improved efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2511_12442
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Global-Lens Transformers: Adaptive Token Mixing for Dynamic Link Prediction
Zou, Tao
Wu, Chengfeng
Liao, Tianxi
Ye, Junchen
Du, Bowen
Machine Learning
Artificial Intelligence
Dynamic graph learning plays a pivotal role in modeling evolving relationships over time, especially for temporal link prediction tasks in domains such as traffic systems, social networks, and recommendation platforms. While Transformer-based models have demonstrated strong performance by capturing long-range temporal dependencies, their reliance on self-attention results in quadratic complexity with respect to sequence length, limiting scalability on high-frequency or large-scale graphs. In this work, we revisit the necessity of self-attention in dynamic graph modeling. Inspired by recent findings that attribute the success of Transformers more to their architectural design than attention itself, we propose GLFormer, a novel attention-free Transformer-style framework for dynamic graphs. GLFormer introduces an adaptive token mixer that performs context-aware local aggregation based on interaction order and time intervals. To capture long-term dependencies, we further design a hierarchical aggregation module that expands the temporal receptive field by stacking local token mixers across layers. Experiments on six widely-used dynamic graph benchmarks show that GLFormer achieves SOTA performance, which reveals that attention-free architectures can match or surpass Transformer baselines in dynamic graph settings with significantly improved efficiency.
title Global-Lens Transformers: Adaptive Token Mixing for Dynamic Link Prediction
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2511.12442