AI Paper Atlas

Research map / Graph learning / GNN architectures

Pareto-Based Heterogeneous Knowledge Distillation for MLPs on Graphs

Wenrui Zhao, Yijun Tian, Zhichao Xu, Yawei Wang, Chuxu Zhang

AAAI 2026 · Graph learning › GNN architectures · approach: Simple and scalable GNNs · 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
distill HGNN teacher into MLP with HGKD instead of training MLP on raw featuresMLP student · IMDB/DBLPMacro-F1: IMDB 20%: 49.63 -> 57.12 (HAN teacher)Improved
HGKD student vs HAN teacher on DBLPHGKD student · DBLPMacro-F1: 91.89 -> 91.21 (20% train)Hurt
HGKD student vs HAN teacher on IMDBHGKD student · IMDBMacro-F1: 63.66 -> 65.06 (80% train)Improved
multi-level HGKD vs GLNN single-level distillation (ablation)MLP student · DBLPMacro-F1: 87.32 -> 92.13 (20% train, best GLNN vs HGKD)Improved
add limited neighbor features (HGKD-N) (ablation)HGKD · IMDBMacro-F1: 57.12 -> 58.52 (HAN, 20%)Improved
add limited neighbor features (HGKD-N) (ablation)HGKD · DBLPMacro-F1: 94.03 -> 93.99 (HAN, 80%)No clear effect
add Pareto-based multi-teacher learning (HGKD+) (ablation)HGKD · IMDBMacro-F1: 60.52 -> 60.50 (40%) vs 57.12 -> 57.76 (20% HAN)Mixed
combine multi-teacher and neighbor info (HGKD-N+) (ablation)HGKD+ · IMDBMacro-F1: 65.06 -> 67.72 (80% vs HGKD-HAN)Improved

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