CauVQ: Causal Vector Quantization for Graph OOD Generalization
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (33) · 28409-28417
DOI: 10.1609/aaai.v40i33.40070 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Weihong Zhang und 3 weiteren Autor:innen.
Abstract (Originalsprache)
Graph Neural Networks (GNNs) perform well on in-distribution data but often fail under out-of-distribution (OOD) shifts due to reliance on spurious patterns. To address this, we propose CauVQ, a causal vector quantization framework that improves OOD generalization by identifying and leveraging invariant substructures that are causally predictive. To construct stable and symbolic graph representations, CauVQ decomposes each input into local substructures and maps them to a discrete codebook of prototypical motifs. This enables consistent and interpretable encoding across diverse graph domains. To isolate the causal substructures, we maximize their mutual information with graph labels and refine their representations using a learnable interaction matrix and a causal attention mechanism. Furthermore, we introduce a counterfactual regularization strategy to enforce prediction stability under substructure perturbations, encouraging the model to focus on truly causal patterns rather than superficial shortcuts. Extensive experiments across standard and OOD benchmarks demonstrate that CauVQ consistently outperforms state-of-the-art baselines in robustness and interpretability. Our framework offers a promising step toward reliable, explainable, and distribution-aware graph learning.
Zitieren
APA
Zhang, W., Bai, L., Du, H., & Yang, X. (2026). CauVQ: Causal Vector Quantization for Graph OOD Generalization. Proceedings of the AAAI Conference on Artificial Intelligence, 40(33), 28409-28417. https://doi.org/10.1609/aaai.v40i33.40070
Vancouver
Zhang W, Bai L, Du H, Yang X. CauVQ: Causal Vector Quantization for Graph OOD Generalization. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(33):28409-28417. doi:10.1609/aaai.v40i33.40070
BibTeX
@article{Zhang2026,
author = {Zhang, Weihong and Bai, Liang and Du, Hangyuan and Yang, Xian},
title = {CauVQ: Causal Vector Quantization for Graph OOD Generalization},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {33},
pages = {28409--28417},
doi = {10.1609/aaai.v40i33.40070}
}
EUVIMED