AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

Scalable Semi-supervised Community Search via Graph Transformer on Attributed Heterogeneous Information Networks

Linlin Ding, Zhaosong Zhao, Mo Li, Yishan Pan, Xin Wang, Renata Borovica-Gajic

AAAI 2026 · Graph learning › Graph self-supervised learning · approach: LLMs and graph foundation models · 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
use SCSAH instead of strongest AHIN baselineSCSAH · DBLPF1 (%): 79.25 -> 87.76Improved
use SCSAH instead of strongest baselineSCSAH · IMDBF1 (%): 66.05 -> 75.34Improved
use SCSAH on MAG where only TransZero also runsSCSAH · MAGF1 (%): 32.59 -> 42.45Improved
add MvSF2Token multi-view subgraph tokens (ablation)SCSAH · IMDB/DBLP/ACMF1/JAC/NMIImproved
add Zoom-Aware Transformer instead of standard Transformer (ablation)SCSAH · IMDB/DBLP/ACMF1/JAC/NMIImproved
add unified contrastive loss (ablation)SCSAH · IMDB/DBLP/ACMaverage performance: 15.2% average drop when removedImproved
add auxiliary node classification and semantic contrastive loss (ablation)SCSAH · IMDB/DBLP/ACMF1/JAC/NMIImproved
set cohesion threshold theta to 0.2 (vs higher values) (ablation)SCSAH · IMDB/DBLPF1/JAC/NMIImproved
use 1-2 Transformer layers instead of more (ablation)SCSAH · IMDB/DBLPF1/JAC/NMIImproved
use subgraph tokens to cut training costSCSAH · IMDB-OAGtraining time: 10.43x faster trainingImproved

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