CondPSE: A Polynomial-Filtered Structural Encoder with Conditional Modulation for Graphs

Hogun Park; Woohyun Lee

REVIEW 3 major objections 5 minor 25 references

CondPSE: A Polynomial-Filtered Structural Encoder with Conditional Modulation for Graphs

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper shows that a learned structural encoder's strong discrimination of synthetic hard graphs does not automatically transfer to better downstream molecular prediction.

desk verdict A useful, honest negative result about learned PSE transfer, but inherited baselines soften the head-to-head conclusion. read the letter →

arxiv 2607.25169 v1 pith:DNYMBCTA submitted 2026-07-28 cs.LG cs.AI

Woohyun Lee , Hogun Park This is my paper · ORCID

classification cs.LGcs.AI

keywords GraphNeuralNetworksPositionalEncodingExpressivityTransferLearningPolynomialFilteringStructuralConditionalModulationGPSE

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper introduces CondPSE, a learned positional/structural encoder that feeds random Gaussian node probes through a bank of learnable polynomial graph filters, creating multiple structural-response branches that are then refined by FiLM-style conditioning using cross-filter, local, and global signals. Pretrained on OGBG-MolPCBA to reconstruct structural and positional targets, CondPSE is frozen and supplied to downstream GNNs. On synthetic isomorphism benchmarks, it far outperforms the previous learned encoder GPSE (CSL accuracy 42.9→97.3%, EXP 68.3→99.9%), but on molecular property prediction it ends up comparable to GPSE, with a ZINC backbone sweep showing no consistent ordering. The paper's central claim is that stronger benchmark-level structural discrimination does not automatically translate into stronger downstream graph-level transfer, and it argues that downstream integration and pretraining-target alignment are likely responsible.

What carries the argument

The load-bearing component is the polynomial graph filter bank: K learnable polynomial filters applied to i.i.d. Gaussian node probes, where each filter is a normalized mixture of adjacency-propagation powers A^i Z. This creates multiple structural-response branches, and each subsequent conditioning block applies FiLM modulation using cross-filter pairwise comparisons, a shared ResGatedGCN local pass, and graph-level statistics. The filter bank converts random probes into topology-sensitive encodings without any eigendecomposition.

What would settle it

Reproduce the full evaluation by running GPSE and CondPSE under identical random seeds, data splits, and CV protocol on CSL, EXP, ZINC, MolHIV, MolPCBA, and PCQM4Mv2-subset, and check whether CondPSE's synthetic gains persist and whether any molecular ordering changes; if CondPSE then clearly surpasses GPSE on a molecular task, the paper's central conclusion would need to be revised.

Watch

Extended reading notes

Core claim

The central discovery is an empirical separation between benchmark-level structural expressivity and downstream transfer for frozen learned PSE encoders: CondPSE, built from a polynomial filter bank and conditional modulation, can almost perfectly discriminate CSL and EXP graphs that 1-WL-bounded message passing cannot, yet those gains do not produce a downstream advantage over GPSE on molecular property prediction. The paper establishes this by freezing the encoder after pretraining and evaluating it under the same downstream protocols as GPSE, then dissecting the result with ablations and a backbone sweep.

Load-bearing premise

The headline comparisons rely on reusing GPSE, LapPE, and RWSE results from the GPSE paper or its public implementation under the assumption that they were generated under CondPSE's exact seed, split, and cross-validation protocol; if that assumption fails, the relative improvements and the 'no better than GPSE' conclusion could shift.

Editorial extensions

If this is right

Synthetic structural-discrimination benchmarks like CSL and EXP are not, on their own, reliable predictors of a frozen PSE encoder's downstream value; future encoder evaluations should report transfer results on multiple real-world tasks.
The same frozen encoding yields different relative performance across downstream backbones (e.g., CondPSE beats GPSE on GCN/GatedGCN/Transformer but loses on GIN/GINE on ZINC), so conclusions about encoder quality should be drawn from a backbone sweep, not a single architecture.
Because the polynomial filter bank accounts for most of the CSL/EXP gain, future PSE encoders may benefit from explicit multi-depth propagation responses rather than stacked message passing only.
Molecular transfer being close between CondPSE and GPSE suggests the bottleneck for frozen learned PSE encoders may lie in pretraining-target alignment and downstream integration rather than in structural expressivity.
The paper's results caution against equating higher expressivity with better generalization: overly fine structural distinctions could reduce the downstream model's ability to generalize within property-similar graph classes.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The paper leaves unexplored whether fine-tuning (rather than freezing) CondPSE would let its strong structural signal translate into molecular gains; a natural extension is to run the downstream backbones with CondPSE weights trainable.
Because the downstream encodings are generated from a single fixed probe realization, the results could be probe-seed dependent; a testable extension is to average over multiple probe realizations when generating encodings and check variance.
The argument that expressivity can hurt generalization points to a measurable prediction: CondPSE encodings should produce larger intra-class variance than GPSE encodings on property-similar molecular graphs; a reader could compute this variance directly from the stored encodings.
The filter-bank design could be reused as a differentiable component inside a downstream GNN rather than as a frozen input encoder; the paper does not test that configuration.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

The core of this paper is a clean, negative finding: a learned PSE encoder that crushes CSL/EXP (97.3% and 99.9%, vs GPSE's 42.9% and 68.3%) does not transfer any better than GPSE on molecular benchmarks. That is worth knowing. The architecture, CondPSE, is a sensible variant of GPSE — polynomial filter bank over Gaussian probes plus FiLM-style conditioning — and the ablations attribute most of the synthetic gain to the filter bank. Ten-run five-fold CV makes the CSL/EXP numbers credible, and the authors are explicit that the molecular results are mixed and that the ZINC sweep is a diagnostic, not a superiority claim.

The soft spot is exactly what the stress-test note flags: every non-CondPSE baseline is inherited from the GPSE paper or its public implementation, not rerun under CondPSE's seed draws. The molecular differences are small enough that a one-percent baseline shift could flip 'without surpassing' into 'surpassing' on one or two datasets. That said, the qualitative message — huge synthetic gains, no clear downstream win — is likely robust, because even if GPSE's CSL were higher, the transfer gap to molecular tasks remains visible. But as written, the head-to-head numbers are not fully self-contained.

There are also smaller gaps: no significance tests on the molecular differences, no code/data release, and the 'target alignment' discussion is explicitly speculative. None of these are fatal, but combined with the inherited baselines, they keep the paper from being a fully closed empirical study. The most useful fix is to rerun GPSE under the same protocol or at least verify the inherited numbers reproduce; the second is to release the probe-seeded encodings.

Who is this for? People working on learned positional/structural encodings who care about whether synthetic expressivity benchmarks predict transfer. It is a solid workshop-level contribution and a legitimate diagnostic. With the baselines verified and code released, it would be publishable at a main conference workshop or a short paper venue. I would send it to a serious referee, but I'd tell them the load is on the comparison. I would not cite it in the next twelve months until the baselines are confirmed.

Referee Report

3 major / 5 minor

Summary. The paper introduces CondPSE, a learned positional/structural encoding (PSE) encoder that applies a bank of K learnable polynomial graph filters to Gaussian node probes and refines the resulting branches with FiLM-style conditioning using cross-filter, local message-passing, and global graph-level signals. CondPSE is pretrained on OGBG-MolPCBA using the same node-level and graph-level pretraining targets as GPSE, then frozen and used as input encoding for downstream GNNs. On synthetic isomorphism benchmarks, CondPSE raises CSL accuracy from 42.9% (GPSE-reported) to 97.3% and EXP from 68.3% to 99.9%. On molecular benchmarks (ZINC, MolHIV, MolPCBA, PCQM4Mv2-subset) with a GraphGPS backbone, CondPSE performs comparably to GPSE without consistently surpassing it. A ZINC backbone sweep shows mixed relative performance. The authors conclude that stronger benchmark-level structural discrimination does not automatically translate into better downstream graph-level transfer.

Significance. If the central result holds, the paper makes a valuable diagnostic contribution: a frozen learned PSE encoder can separate structures that 1-WL-bounded message passing cannot (CSL/EXP), yet this expressivity does not by itself yield a downstream molecular advantage. This challenges the implicit assumption that stronger synthetic structural discrimination should transfer to real-world tasks. The paper is transparent about its limitations, provides full encoder and training configuration tables, and reports multi-seed results with error bars. The ablation study cleanly isolates the contribution of the polynomial filter bank, and the authors explicitly disavow a definitive superiority claim on the backbone sweep. These strengths make the paper a useful reference for the PSE community.

major comments (3)

[A.3, Tables 1–3] All non-CondPSE baselines (No PE, GPSE, LapPE, RWSE) are inherited from the GPSE paper or its public implementation rather than rerun under the paper's exact seed/CV protocol. The headline comparisons — 'raises CSL accuracy from 42.9% to 97.3%' and 'performs comparably to GPSE without surpassing it' — both depend on these externally sourced numbers. Because CondPSE is evaluated with its own seed draws, a small shift in the GPSE baseline (e.g., CSL 42.9±7.9 under another seed protocol) could alter the claimed gains. The paper should rerun the baselines under identical folds, seeds, and downstream configurations, or at minimum provide cross-seed variance and paired comparisons where feasible.
[A.2, Tables 2–3] CondPSE is pretrained on OGBG-MolPCBA alone, whereas GPSE is pretrained on a much larger and more diverse corpus (the GPSE paper uses a multi-dataset pretraining set). The paper does not control for this difference in pretraining data scale/diversity. The molecular transfer results could therefore reflect not only the encoder architecture but also the amount and composition of pretraining data. To support the claim that 'stronger CSL/EXP performance does not translate to downstream advantage,' CondPSE should be pretrained on the same pretraining corpus as GPSE, or the comparison should explicitly account for this confound.
[Tables 2–3] The molecular differences between GPSE and CondPSE are not subjected to any significance test, and several are well within one standard deviation: ZINC MAE 0.0648±0.0030 vs 0.0649±0.0035; PCQM4Mv2-subset MAE 0.1196±0.0004 vs 0.1207±0.0006. The claim that CondPSE is 'without surpassing' GPSE is essentially an acceptance of the null hypothesis. With 10 seeds (and 4 seeds in the sweep), paired tests such as Wilcoxon signed-rank or a bootstrap over seeds are straightforward and would substantiate or qualify the conclusion. Without such tests, the central negative finding is not statistically grounded.

minor comments (5)

[3.1] The term 'NormalSE' is used without explicit definition at first mention. Clarify that it denotes standard i.i.d. Gaussian node probes, and please keep the terminology consistent ('NormalSE' vs. 'Gaussian probes').
[Table 1] The 'w/o filter bank' ablation is not described. Specify exactly what replaces the polynomial filter bank in this variant (e.g., a single linear transformation of A^0 Z, randomly initialized untrained filters, or a different propagation scheme) and report the parameter count if it differs.
[Table 3] The statement 'several gaps are within reported variation' would benefit from concrete numbers. Indicate which differences are beyond one standard deviation and, ideally, add error bars or a significance test for the backbone sweep as well.
[Abstract] The phrasing 'raises CSL accuracy from 42.9% to 97.3% relative to GPSE' could be read as a direct head-to-head experiment. Because the 42.9% value is inherited, consider rephrasing to 'compared to the GPSE-reported value' or 'exceeds the GPSE-reported accuracy.'
[5] The expressivity-generalization trade-off discussion is explicitly untested here, which is appropriate. The references [14,15] are cited as supporting theory; ensure they are publicly available and correctly described, especially [15], which is dated 2026.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: CondPSE's pretraining targets are external, downstream labels are unused during pretraining, and the headline transfer comparison rests on independent benchmark numbers rather than a self-referential derivation.

full rationale

Walking the paper's derivation chain: CondPSE is defined by Eq. (1)-(5) as a polynomial-filtered, FiLM-modulated encoder over Gaussian probes; these equations do not reference downstream labels or evaluation metrics. Pretraining (Section A.2) reconstructs established external targets (ElectrostaticPE, LapPE, RWSE, HKdiagSE, EigVals, CycleGE) on OGBG-MolPCBA, with no downstream property labels used. The CSL/EXP results are presented as a benchmark of whether a frozen learned encoder can recover structural signals that are known to discriminate these graph classes; the paper explicitly discloses that RWSE is among the pretraining targets and that RWSE already attains 100.0 on CSL (Table 1 and Section 4.1). This is an honest representation-quality evaluation, not a circular prediction: GPSE is pretrained on the same targets yet scores 42.9 on CSL, so the CondPSE result has independent content attributable to the architecture, not to the target set. The molecular transfer claim compares CondPSE against baseline numbers inherited from the GPSE paper/public implementation (Section A.3); this is a reproducibility and comparability risk (seed protocol differences), but it is not circularity because the baselines are externally produced, not derived from CondPSE itself, and no parameter is fitted to make the conclusion follow. The Discussion's caveats ('we do not directly measure target-label alignment, so this remains a hypothesis') are explicit uncertainty, not circular reasoning. There is no self-citation chain, no imported uniqueness theorem, and no ansatz smuggled in via citation that reduces a claim to its own input.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim is empirical, so the ledger mainly records hand-chosen architectural hyperparameters and background assumptions about the PSE pretraining paradigm. No new physical or mathematical entities are introduced; the only 'invented' objects are neural-network modules, which have no independent falsifiable existence outside the trained weights.

free parameters (5)

Filter bank size K and polynomial degree d = K=4, d=15
Hand-chosen; controls the number and depth of propagation-response mixtures. No sensitivity analysis over K or d is reported.
Conditioning blocks B = B=6
Hand-chosen number of FiLM modulation blocks; no ablation over B.
Model/probe/output dimensions D = 256
Probe, hidden, and output dimensions all set to 256; capacity choice not motivated by theory.
Downstream adapter dimension d_pe = 128 (CSL), 64 (EXP), 32 (ZINC/MolHIV), 48 (MolPCBA), 64 (PCQM4Mv2-subset)
Per-dataset values inherited from GPSE configs; determines the trade-off between raw feature and encoding space in Eq. 5.
Training schedules (epochs, LR, batch size, dropout) = e.g., 120 pretrain epochs, LR 5e-4, batch 512; downstream varies
Hand-chosen optimizer settings. Not shown to affect the central transfer conclusion, but not swept either.

assumptions (5)

standard math Message-passing GNNs are bounded by the 1-WL graph isomorphism test.
Invoked in Introduction with refs [16,22] to motivate PSE encoders; a standard background theorem.
domain assumption Random Gaussian node probes propagated through the adjacency matrix carry enough structural information to reconstruct the selected PSE targets.
Section 3.1 relies on random-feature expressivity results [1,20] and randomized linear algebra [11,21], but the sufficiency of NormalSE probes for ElectrostaticPE/LapPE/RWSE/HKdiagSE/EigVals/CycleGE is assumed, not proven.
domain assumption The chosen hand-crafted PSE targets are valid supervision for structural discrimination on CSL/EXP.
Pretraining (A.2) uses GPSE's targets; Table 1 shows RWSE alone solves CSL (100%) and near-solves EXP, so gains may largely reflect reconstruction of an already-sufficient target rather than a new structural signal.
domain assumption Frozen encoder plus light downstream adapter plus concatenation (Eq. 5) is a fair and general protocol for comparing learned PSE encoders.
Section 3.3 fixes this integration; if downstream models fine-tuned the encoder or injected the encoding differently, relative results could change.
ad hoc to paper GPSE/LapPE/RWSE baseline numbers from public reports are comparable to CondPSE numbers produced under 'corresponding public GPSE evaluation settings'.
Section A.3 states baselines are taken from the GPSE paper/public implementation, not rerun locally; the headline relative improvements and the 'no better than GPSE' transfer conclusion depend on this comparability.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CondPSE: A Polynomial-Filtered Structural Encoder with Conditional Modulation for Graphs." pith.science (2026). https://pith.science/paper/DNYMBCTA

@misc{pith2026260725169,
  author       = {Pith},
  title        = {Pith review of: CondPSE: A Polynomial-Filtered Structural Encoder with Conditional Modulation for Graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DNYMBCTA}},
  note         = {Machine review of arXiv:2607.25169}
}

read the original abstract

Message-passing graph neural networks are bounded by the 1-WL test and can miss topological structure that distinguishes non-isomorphic graphs. Positional and structural encodings (PSE) inject such topology-derived signals, and learned PSE encoders such as GPSE pretrain a single encoder to produce these signals from random node probes, which can then be frozen and reused as inputs across downstream graph models. We present CondPSE, a learned PSE encoder that applies a learnable polynomial graph filter bank to standard Gaussian node probes and refines the resulting structural-response branches through FiLM-style modulation conditioned on cross-filter, local message-passing, and graph-level signals. CondPSE is pretrained to reconstruct node-level positional/structural targets and graph-level invariants, and is then frozen for use as a downstream input encoding. On synthetic structural-discrimination benchmarks, CondPSE separates graph structures that 1-WL-bounded message passing cannot: it raises CSL accuracy from 42.9% to 97.3% and EXP accuracy from 68.3% to 99.9% relative to GPSE, and ablations show that the polynomial filter bank accounts for most of this gain. On real molecular property prediction, the picture is more limited. With a hybrid local-message-passing/global-attention backbone, CondPSE performs comparably to GPSE without surpassing it, and a ZINC backbone sweep shows no consistent ordering between the two encoders. We report these results and discuss why strong synthetic structural discrimination does not, on its own, yield a downstream advantage for frozen learned PSE encoders, including the role of downstream integration and possible mismatch between structural pretraining targets and molecular property labels.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 1 linked inside Pith

[1]

Ralph Abboud, Ismail Ilkan Ceylan, Martin Grohe, and Thomas Lukasiewicz
[2]

Xavier Bresson and Thomas Laurent. 2017. Residual Gated Graph ConvNets. arXiv preprint arXiv:1711.07553(2017)

arXiv 2017
[3]

Semih Cantürk, Renming Liu, Olivier Lapointe-Gagné, Vincent Létourneau, Guy Wolf, Dominique Beaini, and Ladislav Rampášek. 2024. Graph Positional and Structural Encoder. InProceedings of the 41st International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 235). PMLR, 5533–5566

2024
[4]

Eli Chien, Jianhao Peng, Pan Li, and Olgica Milenkovic. 2021. Adaptive Universal Generalized PageRank Graph Neural Network. InInternational Conference on Learning Representations

2021
[5]

Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. 2016. Convo- lutional Neural Networks on Graphs with Fast Localized Spectral Filtering. In Advances in Neural Information Processing Systems, Vol. 29. 3844–3852

2016
[6]

Joshi, Anh Tuan Luu, Thomas Laurent, Yoshua Bengio, and Xavier Bresson

Vijay Prakash Dwivedi, Chaitanya K. Joshi, Anh Tuan Luu, Thomas Laurent, Yoshua Bengio, and Xavier Bresson. 2023. Benchmarking Graph Neural Networks. Journal of Machine Learning Research24, 43 (2023), 1–48

2023
[7]

Vijay Prakash Dwivedi, Anh Tuan Luu, Thomas Laurent, Yoshua Bengio, and Xavier Bresson. 2022. Graph Neural Networks with Learnable Structural and Po- sitional Representations. InInternational Conference on Learning Representations

2022
[8]

Weihua Hu, Matthias Fey, Hongyu Ren, Maho Nakata, Yuxiao Dong, and Jure Leskovec. 2021. OGB-LSC: A Large-Scale Challenge for Machine Learning on Graphs. InProceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks

2021

Show all 25 references

[9]

Weihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong, Hongyu Ren, Bowen Liu, Michele Catasta, and Jure Leskovec. 2020. Open Graph Benchmark: Datasets for Machine Learning on Graphs. InAdvances in Neural Information Processing Systems, Vol. 33. 22118–22133

2020
[10]

Weihua Hu, Bowen Liu, Joseph Gomes, Marinka Zitnik, Percy Liang, Vijay Pande, and Jure Leskovec. 2020. Strategies for Pre-training Graph Neural Networks. In International Conference on Learning Representations

2020
[11]

Hutchinson

Michael F. Hutchinson. 1990. A stochastic estimator of the trace of the influence matrix for Laplacian smoothing splines.Communications in Statistics: Simulation and Computation19, 2 (1990), 433–450

1990
[12]

Kipf and Max Welling

Thomas N. Kipf and Max Welling. 2017. Semi-Supervised Classification with Graph Convolutional Networks. InInternational Conference on Learning Repre- sentations

2017
[13]

Hamilton, Vincent Létourneau, and Prudencio Tossou

Devin Kreuzer, Dominique Beaini, William L. Hamilton, Vincent Létourneau, and Prudencio Tossou. 2021. Rethinking Graph Transformers with Spectral Attention. InAdvances in Neural Information Processing Systems, Vol. 34. 21618–21629

2021
[14]

Shouheng Li, Floris Geerts, Dongwoo Kim, and Qing Wang. 2025. Towards Bridg- ing Generalization and Expressivity of Graph Neural Networks. InInternational Conference on Learning Representations

2025
[15]

Lutzeyer

Sohir Maskey, Raffaele Paolino, Fabian Jogl, Gitta Kutyniok, and Johannes F. Lutzeyer. 2026. Graph Representational Learning: When Does More Expressivity Hurt Generalization?. InInternational Conference on Learning Representations

2026
[16]

Hamilton, Jan E

Christopher Morris, Martin Ritzert, Matthias Fey, William L. Hamilton, Jan E. Lenssen, Gaurav Rattan, and Martin Grohe. 2019. Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 33. 4602–4609

2019
[17]

Murphy, Balasubramaniam Srinivasan, Vinayak Rao, and Bruno Ribeiro

Ryan L. Murphy, Balasubramaniam Srinivasan, Vinayak Rao, and Bruno Ribeiro
[18]

Ethan Perez, Florian Strub, Harm de Vries, Vincent Dumoulin, and Aaron Courville. 2018. FiLM: Visual Reasoning with a General Conditioning Layer. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 32. 3942– 3951

2018
[19]

Ladislav Rampášek, Michael Galkin, Vijay Prakash Dwivedi, Anh Tuan Luu, Guy Wolf, and Dominique Beaini. 2022. Recipe for a General, Powerful, Scalable Graph Transformer. InAdvances in Neural Information Processing Systems, Vol. 35. 14501–14515

2022
[20]

Ryoma Sato, Makoto Yamada, and Hisashi Kashima. 2021. Random Features Strengthen Graph Neural Networks. InProceedings of the 2021 SIAM International Conference on Data Mining. 333–341

2021
[21]

Shashanka Ubaru, Jie Chen, and Yousef Saad. 2017. Fast estimation of tr(f(A)) via stochastic Lanczos quadrature.SIAM J. Matrix Anal. Appl.38, 4 (2017), 1075–1099

2017
[22]

Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. 2019. How Powerful are Graph Neural Networks?. InInternational Conference on Learning Representa- tions

2019
[23]

Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, and Tie-Yan Liu. 2021. Do Transformers Really Perform Badly for Graph Representation?. InAdvances in Neural Information Processing Systems, Vol. 34. 28877–28888. KDD 2026 Workshop GMLLM, Au...

2021
[2019]

InProceedings of the 36th International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol

Relational Pooling for Graph Representations. InProceedings of the 36th International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 97). PMLR, 4663–4673
[2021]

InProceedings of the Thirtieth International Joint Conference on Artificial Intelligence

The Surprising Power of Graph Neural Networks with Random Node Initialization. InProceedings of the Thirtieth International Joint Conference on Artificial Intelligence. 2112–2118

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Ralph Abboud, Ismail Ilkan Ceylan, Martin Grohe, and Thomas Lukasiewicz

[2] [2]

Xavier Bresson and Thomas Laurent. 2017. Residual Gated Graph ConvNets. arXiv preprint arXiv:1711.07553(2017)

arXiv 2017

[3] [3]

Semih Cantürk, Renming Liu, Olivier Lapointe-Gagné, Vincent Létourneau, Guy Wolf, Dominique Beaini, and Ladislav Rampášek. 2024. Graph Positional and Structural Encoder. InProceedings of the 41st International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 235). PMLR, 5533–5566

2024

[4] [4]

Eli Chien, Jianhao Peng, Pan Li, and Olgica Milenkovic. 2021. Adaptive Universal Generalized PageRank Graph Neural Network. InInternational Conference on Learning Representations

2021

[5] [5]

Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. 2016. Convo- lutional Neural Networks on Graphs with Fast Localized Spectral Filtering. In Advances in Neural Information Processing Systems, Vol. 29. 3844–3852

2016

[6] [6]

Joshi, Anh Tuan Luu, Thomas Laurent, Yoshua Bengio, and Xavier Bresson

Vijay Prakash Dwivedi, Chaitanya K. Joshi, Anh Tuan Luu, Thomas Laurent, Yoshua Bengio, and Xavier Bresson. 2023. Benchmarking Graph Neural Networks. Journal of Machine Learning Research24, 43 (2023), 1–48

2023

[7] [7]

Vijay Prakash Dwivedi, Anh Tuan Luu, Thomas Laurent, Yoshua Bengio, and Xavier Bresson. 2022. Graph Neural Networks with Learnable Structural and Po- sitional Representations. InInternational Conference on Learning Representations

2022

[8] [8]

Weihua Hu, Matthias Fey, Hongyu Ren, Maho Nakata, Yuxiao Dong, and Jure Leskovec. 2021. OGB-LSC: A Large-Scale Challenge for Machine Learning on Graphs. InProceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks

2021