---
id: 20260314-T0-28
title: "图tokenization：Transformer与图桥梁"
title_en: "Graph Tokenization for Transformer Bridging"
url: https://ai.daily.yangsir.net/daily/20260314-T0-28
issue_date: 2026-03-14
publish_date: 2026-03-13T04:00:00.000Z
source_name: "arXiv cs.LG (ML)"
source_url: https://arxiv.org/abs/2603.11099
---

# 图tokenization：Transformer与图桥梁

Graph Tokenization论文提出图结构数据的新tokenization方法。研究将图转换为离散符号序列，使Transformer能直接处理图数据。在OGB数据集上，该方法实现89%的节点分类准确率，比传统GNN高7%。论文arXiv:2603.11099v1为图神经网络与大模型融合提供了关键技术。

## English Version

**Graph Tokenization for Transformer Bridging**

Graph Tokenization paper proposes new method for graph data tokenization. Converts graphs to discrete symbol sequences for direct Transformer processing. Achieves 89% node classification accuracy on OGB, 7% higher than traditional GNNs. arXiv:2603.11099v1.

---

**来源**：[arXiv cs.LG (ML)](https://arxiv.org/abs/2603.11099)

**详情页**：https://ai.daily.yangsir.net/daily/20260314-T0-28

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*