Rethinking Graph Generalization through the Lens of Sharpness-Aware Minimization rfp4136 CCS: Computing methodologies Machine learning
Yang Qiu, Yixiong Zou, Jun Wang · 2026
Read Rethinking Graph Generalization through the Lens of Sharpness-Aware Minimization rfp4136 CCS: Computing methodologies Machine learning by Yang Qiu, Yixiong Zou, Jun Wang (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-2602-08855.txt.
Description
Graph neural networks can fail in a strange way: a test case only slightly different from training data suddenly flips to a wrong answer. Yang Qiu and two co-authors call this Minimal Shift Flip and study it through sharpness-aware minimization, which measures how steep and fragile the loss landscape is around a solution.
They define a Local Robust Radius, the smallest nudge that flips a prediction, and link it to generalization in theory. Training, they find, steadily shrinks that radius, sharpening the landscape. Their fix, an energy-driven augmentation scheme called E2A, breeds fake out-of-distribution samples to widen the margin, and benchmarks show steadier generalization.
Catalog record
| CREATOR | Yang Qiu, Yixiong Zou, Jun Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 60 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-08855