Pith. Sign in

RECORD

Edge-free but Structure-aware: Prototype-Guided Knowledge Distillation from GNNs to MLPs

PaperPDF from arXiv

data files

Paper breakdownThe authors' claims, conditions and measurements, as JSON Signed receiptProof this review has not been altered
X Bluesky LinkedIn Reddit Hacker News
Taiqiang Wu, Zhe Zhao, Jiahao Wang, Xingyu Bai, Lei Wang, Ngai Wong, Yujiu Yang

Machine Learning · Submitted 24 Mar 2023

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim Experimental results on popular graph benchmarks demonstrate the effectiveness and robustness of the proposed PGKD.

how it is stated · asserted 1.00 read the record →

Comment (0) Read on arXiv ↗
Cite ▾
arXiv ID2303.13763 (version 3)
Pith IDVSAZ7XNK, Pith's permanent record of this paper and its review
CitationJiahao Wang, Lei Wang, Ngai Wong, Taiqiang Wu, Xingyu Bai, Yujiu Yang, Zhe Zhao. "Edge-free but Structure-aware: Prototype-Guided Knowledge Distillation from GNNs to MLPs." arXiv:2303.13763 (2023). https://pith.science/paper/VSAZ7XNK
Claim paper

The compiled record

The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Open the compiled record

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 5,557 input tokens.

What the abstract itself states, as probabilities: claims first 0.07 · states limitation 0.11 · human participants 0.02 · controlled comparison 0.20 · randomized assignment 0.01 · states quantitative result 0.04 · compares with prior results 0.40 · mentions code or data availability 0.02.

The abstract, sentence by sentence

arXiv listing abstract v3. Results and the main claim are underlined; .

Distilling high-accuracy Graph Neural Networks (GNNs) to low-latency multilayer perceptions (MLPs) on graph tasks has become a hot research topic. However, conventional MLP learning relies almost exclusively on graph nodes and fails to effectively capture the graph structural information. Previous methods address this issue by processing graph edges into extra inputs for MLPs, but such graph structures may be unavailable for various scenarios. To this end, we propose Prototype-Guided Knowledge Distillation (PGKD), which does not require graph edges (edge-free setting) yet learns structure-aware MLPs. Our insight is to distill graph structural information from GNNs. Specifically, we first employ the class prototypes to analyze the impact of graph structures on GNN teachers, and then design two losses to distill such information from GNNs to MLPs. Experimental results on popular graph benchmarks demonstrate the effectiveness and robustness of the proposed PGKD.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#sentencerolethe authors' own workstates the main conclusionstrength
s0 Distilling high-accuracy Graph Neural Networks (GNNs) to low-latency multilayer perceptions (MLPs) on graph tasks has become a hot research topic. background 1.00 0.05 0.03 asserted 1.00
s1 However, conventional MLP learning relies almost exclusively on graph nodes and fails to effectively capture the graph structural information. gap 0.67 0.07 0.03 asserted 0.99
s2 Previous methods address this issue by processing graph edges into extra inputs for MLPs, but such graph structures may be unavailable for various scenarios. background 0.49 0.04 0.03 hedged possibility 0.81
s3 To this end, we propose Prototype-Guided Knowledge Distillation (PGKD), which does not require graph edges (edge-free setting) yet learns structure-aware MLPs. contribution 0.98 0.87 0.30 asserted 1.00
s4 Our insight is to distill graph structural information from GNNs. contribution 0.68 0.51 0.09 asserted 0.99
s5 Specifically, we first employ the class prototypes to analyze the impact of graph structures on GNN teachers, and then design two losses to distill such information from GNNs to MLPs. method 0.99 0.93 0.07 asserted 0.97
s6 Experimental results on popular graph benchmarks demonstrate the effectiveness and robustness of the proposed PGKD. result 0.99 0.94 0.73 asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.98.

What was examined, and what was not

sources read
  • arXiv listing abstract v3
not examined
  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers
levels run
  • P abstract
not run
  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper
Original abstract

Distilling high-accuracy Graph Neural Networks (GNNs) to low-latency multilayer perceptions (MLPs) on graph tasks has become a hot research topic. However, conventional MLP learning relies almost exclusively on graph nodes and fails to effectively capture the graph structural information. Previous methods address this issue by processing graph edges into extra inputs for MLPs, but such graph structures may be unavailable for various scenarios. To this end, we propose Prototype-Guided Knowledge Distillation (PGKD), which does not require graph edges (edge-free setting) yet learns structure-aware MLPs. Our insight is to distill graph structural information from GNNs. Specifically, we first employ the class prototypes to analyze the impact of graph structures on GNN teachers, and then design two losses to distill such information from GNNs to MLPs. Experimental results on popular graph benchmarks demonstrate the effectiveness and robustness of the proposed PGKD.

Human review

How Pith checks papers → · Full audit of this paper →

How the literature uses it

5 arXiv papers cite this in 6 sentences

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

Papers citing it · 1 sentence without a confident label

Cites it as background 4 papers

What it cites · 3 sentences without a confident label

Uses the method or tool of 2 papers
  • To evaluate the ability on different backbones, we select four popular GNNs as the teacher model: GraphSAGE [ref] , GAT [ref] , GCN [ref] and APPNP [ref] , and perform experiments under both transductive and inductive settings.

    Graph Attention Networks · Implementation · 0.90

  • We design the intra-class loss analogous to InfoNCE [ref] .

    Representation Learning with Contrastive Predictive Coding · Prototype-Guided Knowledge Distillation · 0.93

Uses data or values from 1 paper
  • To evaluate the performance of PGKD, we consider six popular benchmarks, including four homophilous graph datasets, namely, Cora [ref] , Citeseer [ref] , Pubmed [ref] , and A-computer [ref] , and two heterophilous graph datasets, namely, Penn94 [ref] and Twitch-gamer [ref] .

    Pitfalls of Graph Neural Network Evaluation · Datasets · 0.97

Cites as background 2 papers
  • Another line is to treat graph structural information as a regularization term, where the nodes connected by edges should be closer [ref] .

    Graph-MLP: Node Classification without Message Passing in Graph · Distill GNNs to MLPs · 0.83

  • Graph Neural Networks (GNNs) are gaining importance in handling non-Euclidean structural data and have achieved start-of-the-art performance across graph machine learning tasks, particularly for the node classification task [ref] .

    Graph Attention Networks · Introduction · 0.96

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.