AI Paper Atlas

Research map / Graph learning / GNN architectures

Rethinking GNNs and Missing Features: Challenges, Evaluation and a Robust Solution

Francesco Ferrini, Veronica Lachi, Antonio Longa, Bruno Lepri, Akiyoshi Matono, Andrea Passerini, Xin Liu, Manfred Jaeger

ICML 2026 · Graph learning › GNN architectures · approach: GNN generalization theory · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
concatenate node features with binary missingness mask (MIM) in RDLRDL · REL-TRIAL site-successMAE: 0.4235 -> 0.3746Improved
add MIM mask to RDLRDL · REL-EVENT user-attendanceMAE: 0.2628 -> 0.2553Improved
add MIM mask to RDLRDL · REL-TRIAL study-adverseMAE: 43.5251 -> 43.1148Improved
add MIM mask to RDLRDL · REL-F1 driver-top3ROC-AUC: 0.755 -> 0.768Improved
add MIM mask to RDL on nearly complete dataRDL · REL-ARXIV author-publication (0.07% missing)MAE: 0.5033 -> 0.5051No clear effect
add MIM mask to RDLRDL · RelBench runtimeruntime per seed: 1512.0 -> 1526.2 s (REL-EVENT)No clear effect
evaluate on dense low-dimensional datasets instead of sparse bag-of-words benchmarksexisting GNN missing-feature methods · Cora, Citeseer, PubMed vs Synthetic, Air, Electric, TadpoleF1 vs missingnessHurt
increase missingness rate on sparse benchmarksall GNN methods · Cora, Citeseer, PubMedF1No clear effect
switch missingness mechanism from uniform to feature-dependent MNARGOODIE, FairAC · Synthetic / ElectricAUC of F1: GOODIE 0.771 -> 0.587 (CD-MNAR); FairAC 0.870 (Electric S-MCAR) vs 0.641 (Synthetic FD-MNAR)Hurt
train on FD-MNAR 50% and test on U-MCAR (distribution shift in missingness)GNN missing-feature methods · R2 regimeF1Hurt
use GNNmim versus other missing-feature GNNsGNN with missing node features · all missingness mechanisms and datasetsAUC of F1Improved
use GNNmim versus MICE, MLP+MIM and XGBoost under CD-MNARGNNmim · CD-MNAR (Appendix L)performanceImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy