arXiv cs.AI· Asiri Wijesinghe, Sevvandi Kandanaarachchi, Daniel M. Steinberg, Cheng Soon Ong·· 14 小时前AI 评分22
Flowette：用 Graphette 先验做流匹配的图生成框架
Flowette: Flow Matching with Graphette Priors for Graph Generation
AI 导读
Flowette 是一个连续流匹配图生成框架，用基于图神经网络的 Transformer 学习带节点与边属性的图表示速度场，并提出 graphettes 这一概率图结构模型族，通过受控结构编辑泛化 graphon 以刻画环、星、树等模体。模型结合最优传输耦合与端点感知、化学感知正则化，在合成与分子基准上多项指标达到 SOTA，消融实验分离了结构先验、耦合与正则项的贡献。
正文 · 原文
Abstract:We study generative modeling of graphs with recurring subgraph motifs. We propose Flowette, a continuous flow matching framework that employs a graph neural network-based transformer to learn a velocity field over graph representations with node and edge attributes. Our model promotes topology-aware alignment through optimal transport-based coupling and preserves structural fidelity through endpoint-aware and chemistry-aware regularisation. To incorporate domain-driven structural priors, we introduce graphettes, a new probabilistic family of graph structure models that generalize graphons via controlled structural edits for motifs such as rings, stars, and trees. We theoretically analyze the coupling, invariance, and structural properties of the framework, evaluate it on synthetic and molecular benchmarks, and isolate the contributions of the structural prior, the optimal-transport coupling, and the regularisation terms through controlled ablations. Flowette achieves competitive performance overall, attaining state-of-the-art results on several metrics across multiple benchmarks, highlighting the effectiveness of combining structural priors with flow-based training for modeling complex graph distributions.
| Comments: | 61 Pages |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2602.23566 [cs.LG] |
| (or arXiv:2602.23566v4 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2602.23566 arXiv-issued DOI via DataCite |
|
| Journal reference: | 40th Conference on Neural Information Processing Systems (NeurIPS 2026) |
Submission history
From: Asiri Wijesinghe [view email]
[v1]
Fri, 27 Feb 2026 00:22:21 UTC (13,421 KB)
[v2]
Fri, 15 May 2026 05:40:18 UTC (14,114 KB)
[v3]
Mon, 18 May 2026 02:44:07 UTC (14,114 KB)
[v4]
Mon, 26 Oct 2026 00:40:01 UTC (16,392 KB)
来源：arXiv cs.AI · arxiv.org