Towards OOD Generalization in Dynamic Graphs via Causal Invariant Learning
Xinxun Zhang, Pengfei Jiao, Mengzhou Gao, Tianpeng Li, Xuan Guo · 2026
Read Towards OOD Generalization in Dynamic Graphs via Causal Invariant Learning by Xinxun Zhang, Pengfei Jiao, Mengzhou Gao, Tianpeng Li, Xuan Guo (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-2603-01626.txt.
Description
Xinxun Zhang, Pengfei Jiao, Mengzhou Gao, and colleagues help graph models cope when data patterns shift over time. Their 2026 work studies dynamic graphs, where nodes and links evolve, in fields from social feeds to traffic. Most models assume test data matches training data, yet real streams drift in ways that break that match.
Their model, DyCIL, hunts for steady causal patterns inside the changing graph. A generator first pulls out dynamic subgraphs that carry the causal core of each change. An attention module then reads how those core patterns evolve through space and time. A third part invents varied shift settings, so training sees more kinds of drift. Tests on real and synthetic dynamic graphs beat top baselines under distribution shift.
Catalog record
| CREATOR | Xinxun Zhang, Pengfei Jiao, Mengzhou Gao, Tianpeng Li, Xuan Guo |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69872 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.01626