![[2602.19392] Spiking Graph Predictive Coding for Reliable OOD Generalization](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19392] Spiking Graph Predictive Coding for Reliable OOD Generalization
Summary
The paper introduces Spiking Graph Predictive Coding (SIGHT), a novel approach to enhance out-of-distribution (OOD) generalization in graph neural networks (GNNs) by improving uncertainty estimation and interpretability.
Why It Matters
As machine learning models face challenges with OOD shifts, SIGHT offers a solution that enhances predictive accuracy and trustworthiness in dynamic environments, crucial for applications in high-stakes settings like Web4Good.
Key Takeaways
- SIGHT is designed to improve uncertainty estimation in GNNs.
- The method enables models to identify and correct internal mismatches in predictions.
- SIGHT enhances both predictive accuracy and interpretability across various benchmarks.
Computer Science > Machine Learning arXiv:2602.19392 (cs) [Submitted on 22 Feb 2026] Title:Spiking Graph Predictive Coding for Reliable OOD Generalization Authors:Jing Ren, Jiapeng Du, Bowen Li, Ziqi Xu, Xin Zheng, Hong Jia, Suyu Ma, Xiwei Xu, Feng Xia View a PDF of the paper titled Spiking Graph Predictive Coding for Reliable OOD Generalization, by Jing Ren and 8 other authors View PDF HTML (experimental) Abstract:Graphs provide a powerful basis for modeling Web-based relational data, with expressive GNNs to support the effective learning in dynamic web environments. However, real-world deployment is hindered by pervasive out-of-distribution (OOD) shifts, where evolving user activity and changing content semantics alter feature distributions and labeling criteria. These shifts often lead to unstable or overconfident predictions, undermining the trustworthiness required for Web4Good applications. Achieving reliable OOD generalization demands principled and interpretable uncertainty estimation; however, existing methods are largely post-hoc, insensitive to distribution shifts, and unable to explain where uncertainty arises especially in high-stakes settings. To address these limitations, we introduce SpIking GrapH predicTive coding (SIGHT), an uncertainty-aware plug-in graph learning module for reliable OOD Generalization. SIGHT performs iterative, error-driven correction over spiking graph states, enabling models to expose internal mismatch signals that reveal where predic...
