Papers › Recipe for a General, Powerful, Scalable Graph Transformer

Recipe for a General, Powerful, Scalable Graph Transformer

25 May 2022arXiv:2205.12454archive 2025-07-28

Ladislav Rampášek, Mikhail Galkin, Vijay Prakash Dwivedi, Anh Tuan Luu, Guy Wolf, Dominique Beaini

We propose a recipe on how to build a general, powerful, scalable (GPS) graph Transformer with linear complexity and state-of-the-art results on a diverse set of benchmarks. Graph Transformers (GTs) have gained popularity in the field of graph representation learning with a variety of recent publications but they lack a common foundation about what constitutes a good positional or structural encoding, and what differentiates them. In this paper, we summarize the different types of encodings with a clearer definition and categorize them as being local, global or relative. The prior GTs are constrained to small graphs with a few hundred nodes, here we propose the first architecture with a complexity linear in the number of nodes and edges O(N+E) by decoupling the local real-edge aggregation from the fully-connected Transformer. We argue that this decoupling does not negatively affect the expressivity, with our architecture being a universal function approximator on graphs. Our GPS recipe consists of choosing 3 main ingredients: (i) positional/structural encoding, (ii) local message-passing mechanism, and (iii) global attention mechanism. We provide a modular framework GraphGPS that supports multiple types of encodings and that provides efficiency and scalability both in small and large graphs. We test our architecture on 16 benchmarks and show highly competitive results in all of them, show-casing the empirical benefits gained by the modularity and the combination of different strategies.

PaperPDFCodeCode Syntology ranMethod as code

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

No sample from the archive's official repository (rampasek/GraphGPS) is linked to this paper; the 21 samples below come from 1 community repository. Syntology harvested code from rampasek/GraphGPS for another paper.

rampasek/GraphGPS: Syntology's graph links these samples to Graph Convolutions Enrich the Self-Attention in Transformers! (arXiv:2312.04234; that paper's own run record: 22 ran (of which 0 constructed an object rather than computing a result; 11 with no instrument failure: 2 honoured, 5 violated, 4 with no contract checked; 11 where Syntology's instrument failed) · 7 unverified); nothing checks that they implement this paper's method.

Syntology 21 code samples harvested from 1 repository linked to this paper: 3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 2 where Syntology's instrument failed) · 18 unverified. Where Syntology's instrument failed: 1 where our contract draft was wrong; 1 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: community (archive-listed): 21 samples from 1 repository, 3 ran (1 with no instrument failure; 2 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 3 of its repositories that Syntology harvested for other papers:

Levels written between 2026-09-22 and 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

rampasek/GraphGPS officialmentioned in papermentioned on GitHubpytorchMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Graph Convolutions Enrich the Self-Attention in Transformers! (arXiv:2312.04234; that paper's own run record: 22 ran (of which 0 constructed an object rather than computing a result; 11 with no instrument failure: 2 honoured, 5 violated, 4 with no contract checked; 11 where Syntology's instrument failed) · 7 unverified); nothing checks that they implement this paper's method. report
graphcore/ogb-lsc-pcqm4mv2 mentioned on GitHubtfSyntology: 21 samples, 3 ran (2 instrument-failed) · 18 unverified report
hamed1375/exphormer mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Exphormer: Sparse Transformers for Graphs (arXiv:2303.06147; that paper's own run record: 1 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
linusbao/MoSE mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Graph Inductive Biases in Transformers without Message Passing (arXiv:2305.17589; that paper's own run record: 6 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 1 honoured, 0 violated, 1 with no contract checked; 4 where Syntology's instrument failed) · 6 unverified) and Homomorphism Counts as Structural Encodings for Graph Learning (arXiv:2410.18676; that paper's own run record: 2 ran (of which 1 constructed an object rather than computing a result; 2 with no instrument failure: 1 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

21 samples harvested: 3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 2 where Syntology's instrument failed) · 18 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

1ran · our draft was wrong
1ran; the contract check could not run
1ran
18unverified

Licence: 0 of the 21 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from graphcore/ogb-lsc-pcqm4mv2. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

graphcore/ogb-lsc-pcqm4mv2 community (archive-listed) 21 samples, 3 ran (2 instrument-failed) · 18 unverified
_scatter_min model/hybrid/layers.py ran · our draft was wrong MIT (permissive) · 23d181463d883ce2 · report fixture: signature data:EagerTensor/2/float, indices:EagerTensor/2/int, num_segments:int/0/int, gather_scatter_method:str/0/str, backwards_mode:str/0/str · returned EagerTensor · level recorded 2026-09-24
_scatter_sum model/hybrid/layers.py ran; the contract check could not run (RAISES) MIT (permissive) · 734d560a1dc6e70b · report fixture: signature data:EagerTensor/3/float, indices:EagerTensor/2/int, num_segments:int/0/int, gather_scatter_method:str/0/str · returned EagerTensor · level recorded 2026-09-24
generate_dropout_seed model/hybrid/layers.py ran MIT (permissive) · b9465ab1ad9e2278 · report fixture: returned EagerTensor · level recorded 2026-09-22
BiasedSelfAttentionLayer model/hybrid/layers.py unverified MIT (permissive) · e2ab63adef0914be · report
FFNLayer model/hybrid/layers.py unverified MIT (permissive) · c84df930a004726a · report
GPSLayer model/hybrid/layers.py unverified MIT (permissive) · 6d03ea56f29b99a2 · report
GenericAggregator model/hybrid/layers.py unverified MIT (permissive) · 154587abcabb598b · report
GraphDropout model/hybrid/layers.py unverified MIT (permissive) · 1fd88eabe2bf5a70 · report
InteractionNetworkLayer model/hybrid/layers.py unverified MIT (permissive) · 971b7db74f2673f3 · report
ResnetMultiplier model/hybrid/layers.py unverified MIT (permissive) · 52b6ed371f4e7c09 · report
_gather model/hybrid/layers.py unverified MIT (permissive) · 401bcf859fc41772 · report
_scatter_count model/hybrid/layers.py unverified MIT (permissive) · 5dda6b2e7b928bd7 · report
_scatter_max model/hybrid/layers.py unverified MIT (permissive) · 14e69fb09fcdefa4 · report
_scatter_mean model/hybrid/layers.py unverified MIT (permissive) · 13e732c6b1e941a0 · report
_scatter_mean_and_var model/hybrid/layers.py unverified MIT (permissive) · 1dbf2926681fecd6 · report
_scatter_softmax model/hybrid/layers.py unverified MIT (permissive) · 8cdb18106afd6290 · report
_scatter_sqrtN model/hybrid/layers.py unverified MIT (permissive) · 7f92923b18f2f819 · report
_scatter_std model/hybrid/layers.py unverified MIT (permissive) · 1e17cdf620f5b2c6 · report
_scatter_var model/hybrid/layers.py unverified MIT (permissive) · eba3c0cda10371e3 · report
call_outlined_function model/hybrid/layers.py unverified MIT (permissive) · 8584b125423d1de0 · report
gather in 75 papers model/hybrid/layers.py unverified MIT (permissive) · df3a80c6ac89079c · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it.

1 assembly on this paper: 1 harvested, 0 generated.

harvested Assembly for GRAPHGPS (graph method: GRAPHGPS). 21 bricks; individually on a fixture: 3 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 2 where Syntology's instrument failed) · 18 unverified. The assembly has not been run end-to-end.

  1. GPSLayer entry · level 0 · unverified · 6d03ea56f29b99a2
  2. BiasedSelfAttentionLayer model · level 0 · unverified · e2ab63adef0914be
  3. FFNLayer model · level 0 · unverified · c84df930a004726a
  4. GraphDropout helper · level 0 · unverified · 1fd88eabe2bf5a70
  5. InteractionNetworkLayer helper · level 0 · unverified · 971b7db74f2673f3
  6. gather helper · level 0 · unverified · df3a80c6ac89079c
  7. GenericAggregator helper · level 0 · unverified · 154587abcabb598b
  8. ResnetMultiplier helper · level 0 · unverified · 52b6ed371f4e7c09
  9. generate_dropout_seed helper · level 1 · ran · b9465ab1ad9e2278
  10. _gather helper · level 0 · unverified · 401bcf859fc41772
  11. call_outlined_function helper · level 0 · unverified · 8584b125423d1de0
  12. _scatter_max helper · level 0 · unverified · 14e69fb09fcdefa4
  13. _scatter_mean helper · level 0 · unverified · 13e732c6b1e941a0
  14. _scatter_mean_and_var helper · level 0 · unverified · 1dbf2926681fecd6
  15. _scatter_min helper · level 1 · ran · our draft was wrong · 23d181463d883ce2
  16. _scatter_softmax helper · level 0 · unverified · 8cdb18106afd6290
  17. _scatter_sqrtN helper · level 0 · unverified · 7f92923b18f2f819
  18. _scatter_std helper · level 0 · unverified · 1e17cdf620f5b2c6
  19. _scatter_sum helper · level 1 · ran; the contract check could not run (RAISES) · 734d560a1dc6e70b
  20. _scatter_var helper · level 0 · unverified · eba3c0cda10371e3
  21. _scatter_count helper · level 0 · unverified · 5dda6b2e7b928bd7

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2205.12454") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

Graph ClassificationGraph Property PredictionGraph RegressionGraph Representation LearningLink PredictionMolecular Property PredictionNode ClassificationRepresentation Learning

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Graph Classification CIFAR10 100k GPS Accuracy (%) 72.298 #11 of 20 Archive leaderboard report
Graph Classification ENZYMES GraphGPS Accuracy 78.667±4.625 #2 of 54 Archive leaderboard report
Graph Classification IMDb-B GraphGPS Accuracy 79.250±3.096 #10 of 51 Archive leaderboard report
Graph Classification MNIST GPS Accuracy 98.05 #12 of 13 Archive leaderboard report
Graph Classification MalNet-Tiny GPS Accuracy 93.36 ± 0.6 #4 of 4 Archive leaderboard report
Graph Classification NCI1 GraphGPS Accuracy 85.110±1.423 #13 of 69 Archive leaderboard report
Graph Classification NCI109 GraphGPS Accuracy 81.256±0.501 #20 of 38 Archive leaderboard report
Graph Classification PROTEINS GraphGPS Accuracy 77.143±1.494 #35 of 103 Archive leaderboard report
Graph Classification Peptides-func GPS AP 0.6535±0.0041 #30 of 44 Archive leaderboard report
Graph Property Prediction ogbg-code2 GPS Ext. data No #6 of 21 Archive leaderboard report
Graph Property Prediction ogbg-code2 GPS Number of params 12454066 #6 of 21 Archive leaderboard report
Graph Property Prediction ogbg-code2 GPS Test F1 score 0.1894 #6 of 21 Archive leaderboard report
Graph Property Prediction ogbg-code2 GPS Validation F1 score 0.1739 ± 0.001 #6 of 21 Archive leaderboard report
Graph Property Prediction ogbg-molhiv GPS Ext. data No #27 of 43 Archive leaderboard report
Graph Property Prediction ogbg-molhiv GPS Number of params 558625 #27 of 43 Archive leaderboard report
Graph Property Prediction ogbg-molhiv GPS Test ROC-AUC 0.7880 #27 of 43 Archive leaderboard report
Graph Property Prediction ogbg-molhiv GPS Validation ROC-AUC 0.8255 ± 0.0092 #27 of 43 Archive leaderboard report
Graph Property Prediction ogbg-molpcba GPS Ext. data No #20 of 36 Archive leaderboard report
Graph Property Prediction ogbg-molpcba GPS Number of params 9744496 #20 of 36 Archive leaderboard report
Graph Property Prediction ogbg-molpcba GPS Test AP 0.2907 #20 of 36 Archive leaderboard report
Graph Property Prediction ogbg-molpcba GPS Validation AP 0.3015 ± 0.0038 #20 of 36 Archive leaderboard report
Graph Property Prediction ogbg-ppa GPS Ext. data No #6 of 18 Archive leaderboard report
Graph Property Prediction ogbg-ppa GPS Number of params 3434533 #6 of 18 Archive leaderboard report
Graph Property Prediction ogbg-ppa GPS Test Accuracy 0.8015 #6 of 18 Archive leaderboard report
Graph Property Prediction ogbg-ppa GPS Validation Accuracy 0.7556 ± 0.0027 #6 of 18 Archive leaderboard report
Graph Regression Lipophilicity GraphGPS R2 0.790±0.004 #8 of 23 Archive leaderboard report
Graph Regression Lipophilicity GraphGPS RMSE 0.579±0.006 #8 of 23 Archive leaderboard report
Graph Regression PCQM4Mv2-LSC GPS Test MAE 0.0862 #9 of 20 Archive leaderboard report
Graph Regression PCQM4Mv2-LSC GPS Validation MAE 0.0852 #9 of 20 Archive leaderboard report
Graph Regression Peptides-struct GPS MAE 0.2500±0.0005 #22 of 39 Archive leaderboard report
Graph Regression ZINC GPS MAE 0.070 ± 0.002 #9 of 27 Archive leaderboard report
Graph Regression ZINC GINE MAE 0.070 ± 0.004 #10 of 27 Archive leaderboard report
Graph Regression ZINC-500k GPS MAE 0.070 #12 of 36 Archive leaderboard report
Graph Regression ZINC-full GraphGPS Test MAE 0.024±0.007 #7 of 19 Archive leaderboard report
Link Prediction PCQM-Contact GPS MRR 0.3337±0.0006 #17 of 18 Archive leaderboard report
Molecular Property Prediction ESOL GraphGPS R2 0.911±0.003 #9 of 20 Archive leaderboard report
Molecular Property Prediction ESOL GraphGPS RMSE 0.613±0.010 #9 of 20 Archive leaderboard report
Molecular Property Prediction FreeSolv GraphGPS R2 0.861±0.037 #12 of 22 Archive leaderboard report
Molecular Property Prediction FreeSolv GraphGPS RMSE 1.462±0.188 #12 of 22 Archive leaderboard report
Node Classification CLUSTER GPS Accuracy 77.95 #9 of 12 Archive leaderboard report
Node Classification COCO-SP GPS macro F1 0.3412±0.0044 #7 of 19 Archive leaderboard report
Node Classification PATTERN GPS Accuracy 86.685 #9 of 11 Archive leaderboard report
Node Classification PascalVOC-SP GPS macro F1 0.3748±0.0109 #8 of 21 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

GPSGTSGraph TransformerLapEigenLaplacian PEMPNNPerformerTransformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections