arXiv：cs.LG· Stefano Carotti, Marco Pacini, Alessio Gravina, Davide Bacciu, Bruno Lepri, Sebastiano Bontorin·· 2 天前AI 评分34
Graph Hierarchical Recurrence：面向长程泛化的图层次递归框架
Graph Hierarchical Recurrence for Long-Range Generalization
AI 导读
研究者提出 Graph Hierarchical Recurrence（GHR）框架，同时在输入图和池化层次抽象上运行，以缓解图神经网络与 Graph Transformer 在远距离图区域关联预测上的局限。GHR 能持续增强所有测试的消息传递骨干网络，在超出训练交互距离的 out-of-range 泛化场景中提升尤为明显。在多项长程基准上，GHR 取得 SOTA 或具竞争力的结果。
正文
Abstract:Graph Neural Networks and Graph Transformers have become central to graph learning, combining expressive representation learning with sample-efficient inductive biases. Yet they remain fundamentally limited when predictions depend on correlations between distant graph regions. We address this limitation with Graph Hierarchical Recurrence (GHR), a novel framework that jointly operates on the input graph and a pooled hierarchical abstraction. We also show that existing models degrade more sharply under out-of-range generalization, where test instances require interactions across distances exceeding those observed during training. Despite its minimal design, GHR consistently strengthens every tested message-passing backbone, yielding robust performance on long-range dependencies and particularly pronounced gains in out-of-range regimes. Across a broad suite of long-range benchmarks, GHR achieves state-of-the-art or competitive results on multiple tasks, establishing hierarchical recurrence as an effective mechanism for extending graph models beyond their observed interaction range.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.18387 [cs.LG] |
| (or arXiv:2605.18387v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.18387 arXiv-issued DOI via DataCite |
Submission history
From: Sebastiano Bontorin [view email]
[v1]
Mon, 18 May 2026 13:31:21 UTC (953 KB)
[v2]
Thu, 1 Oct 2026 15:02:35 UTC (927 KB)
来源：arXiv：cs.LG · arxiv.org