AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

Unifying Graph Contrastive Learning via Graph Message Augmentation

Ziyan Zhang, Bo Jiang, Jin Tang, Bin Luo

TPAMI 2025 · Graph learning › Graph self-supervised learning · approach: Augmentation-free graph learning · 4 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
replace node dropping with message dropping (ablation)GMCL · Coraaccuracy: 83.44 -> 84.30Improved
replace node perturbation with message perturbation (ablation)GMCL · MUTAGaccuracy: 90.96 -> 93.10Improved
replace node dropping with message dropping (ablation)GMCL · MUTAGaccuracy: 91.49 -> 92.08Improved
replace edge dropping with message dropping (ablation)GMCL · IMDB-Baccuracy: 74.20 -> 75.20Improved
replace node dropping with message mixup (ablation)GMCL · REDDIT-Baccuracy: 90.35 -> 92.50Improved
replace edge perturbation with message mixup (ablation)GMCL · Coraaccuracy: 81.72 -> 84.88Improved
replace subgraph sampling with message dropping (ablation)GMCL · MUVROC-AUC: 77.18 -> 78.03Improved
add contrastive pre-training on ChEMBL (GMCL-M vs no pretrain)GIN · ClinToxROC-AUC: 58.0 -> 83.37Improved
add contrastive pre-training on ChEMBL (GMCL-M vs no pretrain)GIN · ToxCastROC-AUC: 63.4 -> 64.12Improved
use GMCL-M instead of AutoGCLGIN · COLLAB (unsupervised)accuracy: 70.12 -> 75.88Improved
use GMCL-M instead of AutoGCLGIN · PROTEINS (unsupervised)accuracy: 77.57 -> 77.44Hurt
use GMCL-M instead of GraphCLGIN · DD (unsupervised)accuracy: 78.62 -> 78.36No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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