Hyperbolic Graph Neural Networks Under the Microscope: The Role of Geometry–Task Alignment
Dionisia Naddeo, Jonas Linkerhägner, Nicola Toschi, Geri Skenderi, Veronica Lachi · 2026
Read Hyperbolic Graph Neural Networks Under the Microscope: The Role of Geometry–Task Alignment by Dionisia Naddeo, Jonas Linkerhägner, Nicola Toschi, Geri Skenderi, Veronica Lachi (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-01828.txt.
Description
Dionisia Naddeo and colleagues question hyperbolic networks, asking whether task and input graph metrics match. Hyperbolic space suits trees, and they add a geometry task alignment condition: does the target metric follow the input metric? Theory and tests show these networks recover low distortion representations on regression, with bias helping when metric structure must be preserved. On link prediction, an aligned task, they gain an advantage, while on node classification, unaligned, the advantage disappears. So the question moves from whether the graph is hyperbolic to whether the task aligns.
Catalog record
| CREATOR | Dionisia Naddeo, Jonas Linkerhägner, Nicola Toschi, Geri Skenderi, Veronica Lachi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 111564 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.01828