AI Paper Atlas

Research map / Graph learning / GNN architectures

Learning from Historical Activations in Graph Neural Networks

Yaniv Galron, Hadar Sinai, Haggai Maron, Moshe Eliasof

ICLR 2026 · Graph learning › GNN architectures · approach: Simple and scalable GNNs · 2 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
add signed division-by-sum normalization of layer-wise attention (vs removing it) (ablation)HistoGraph · PROTEINSaccuracy (%): 74.45 -> 97.80Improved
add layer-wise attention over historical activations (ablation)HistoGraph · PROTEINSaccuracy (%): 78.61 -> 97.80Improved
add node-wise self-attention at readout (ablation)HistoGraph · PROTEINSaccuracy (%): 80.78 -> 97.80Improved
add HistoGraph readout to deep GCNGCN · Cora, 64 layersaccuracy (%): 28.7 -> 77.5Improved
add HistoGraph readout to shallow GCNGCN · Pubmed, 2 layersaccuracy (%): 79.0 -> 78.9No clear effect
HistoGraph vs best baseline pooling on TU with 5-layer GINDKEPool · RDT-Baccuracy (%): 95.0 -> 93.4Hurt
HistoGraph vs best baseline pooling on OGB molhivDKEPool · molhivROC-AUC (%): 78.65 -> 77.81Hurt
HistoGraph vs best baseline on molbbbpDKEPool · molbbbpROC-AUC (%): 69.73 -> 72.02Improved
apply HistoGraph as frozen head (FT) instead of MeanPoolpretrained GIN · IMDB-Baccuracy (%): 76.0 -> 94.0Improved
joint full fine-tuning instead of frozen headHistoGraph post-processing · PTCaccuracy (%): 85.7 -> 97.1Improved
frozen-head fine-tune vs end-to-end HistoGraph trainingHistoGraph · PROTEINSaccuracy (%): 97.3 (FT) vs 97.8 (end-to-end)No clear effect
train HistoGraph end-to-end with the backboneHistoGraph · molhiv 32-layer GCNtraining time per epoch (s): 41.27 -> 60.34Hurt

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