Neural Message Passing for Quantum Chemistry

Justin Gilmer, Samuel S. Schoenholz, Patrick F. Riley, Oriol Vinyals, George E. Dahl · ICML / arXiv · 2017 · chemistry · record gilmer2017-message-passing
tag layer · index · concepts · authors · עברית

Where this record's text came from. faceted 2026-08 by reading the locally held text (pdftext); every evidence string verified as an exact substring of that text before merging. The paper is unusual in that a large part of its contribution is a reformulation of eight named prior models (Duvenaud et al. 2015, Li et al. 2016, Battaglia et al. 2016, Kearnes et al. 2016, Schütt et al. 2017, Bruna et al. 2013, Defferrard et al. 2016, Kipf & Welling 2016) into one framework — a many-to-one 'unifies' relation that the four-type relation vocabulary cannot express, so only three explicit build-on/extend statements are recorded. The paper's critique of Duvenaud et al. (2015) ('unable to identify correlations between edge states and node states') was left out because 'contradicts' would misdescribe it: it is an analytic observation about a model's expressiveness, not a disputed empirical claim. Results reported are single train/validation/test splits with a 50-trial random hyperparameter search; no replication or repeated-split variance is reported. 2026-08-14 and 2026-08-25 schema-0.3 passes: every quote verified as an exact substring of the staged text and anchored at merge time; concept terms verified present; every edge endpoint an existing node. Update, 15 September 2026: EXCEPTION, measured with the merge gate's own predicate (Paper.locate): 2 quote(s) on this record are not found as one exact string in the current staged text - content_tags.statistics[0].evidence, content_tags.statistics[1].evidence. The sentence above does not hold for them as written. Where a row records why - the quote is grounded in another copy of the paper, joins two passages with an ellipsis, or differs from the staged text by extraction noise - that note stays on the row.

The quality of the text these quotes were cut from. symbol-font-not-extractable: 29 '(cid:NNN)' glyph-index markers remain in a staged text of 54,451 characters, where the PDF's embedded font had no ToUnicode map. THE QUOTES ON THIS RECORD ARE THE PAPER'S AND THE STAGED COPY IS THE DAMAGED ONE - which is the opposite of the usual case and is why it went unnoticed: one of its 57 live quotes is not an exact substring of this text, and it was located in it by aligning the quote's WORD SEQUENCE under exactly this damage - the words are there, the characters inside them are not. Not repaired in place: editing the staged text moves every offset after the edit and re-hashes every anchor on the record. Measured at 113aae60fefc3eb1, re-measured unchanged at 54b63482778ddf56.

1 quote on this page breaks a word across a line. The PDF this text was extracted from hyphenated a word at the end of a printed line — adversar- ial, repre- sentations — and the extractor kept the hyphen and the space. The quotes keep them because a quote is reproduced exactly or it proves nothing: the break is the typesetter’s, not the author’s. A search of this page will not find the affected words.

1 quote on this page was cut from a different copy of this article. This project holds two extractions of this paper: the plain text every other quote here is anchored in, and the publisher’s own HTML edition, which we are not licensed to republish. It is verbatim in the second and cannot be found in the first — not because the tagging is wrong, but because the extraction damaged the passage: a page number dropped into the middle of a number, a formula moved out of its sentence, a ligature where two letters belong. The record names the copy for each one. The quote is the paper’s; the extraction is not.

How to read this page. The article itself is not reproduced here — this paper is not cleared for redistribution, so what follows is the tag layer alone. The evidence sentences are withheld with the article. The table at the foot lists every tag the record carries; each row names the facet it belongs to, and the whose words column says whether the tag records the paper's own wording or the tagger's sentence about the paper. Every tag links to each other paper in the corpus carrying it. The machine-readable edition of the same thing is the TEI file beside this one, where the tags live in standOff, and the article is at its publisher via the link below. When we have changed our mind. Some entries in the table begin with a word in capitals and the reason for it. EVIDENCE CORRECTED — the finding still stands, but we had quoted the wrong sentence from the paper, and the quotation has been replaced with the right one. RETRACTED — the entry itself was wrong; it stays on the page, marked, with the reason, rather than being deleted. WITHDRAWN AND REPLACED — the entry was wrong, and a corrected version of it appears in the entries below. Each mark carries its own reason, in brackets, because the reasons differ from one entry to the next. An entry we have WITHDRAWN is removed from this page and kept in the record, with its reason. A line reading Audit and a date is a later re-reading of the entry against its whole sentence: DOUBTFUL means it was kept, and the reason says what is loose about it. Quotation marks inside the tagger's sentence. Where a tag is the tagger's own sentence, a phrase in quotation marks inside it is borrowed from the paper — the tagger's wording around it is not. The paper's own sentence is the evidence quote for that row; on this page it is withheld with the article, and the article is at its publisher. Nothing inside quotation marks anywhere on this page is ours: if it is quoted, it is the paper's. Whose words a concept carries. A concept entry records a term the paper itself uses, so the term, the sense given for it, and its evidence are the paper's own words, quoted where the paper fixes them. This is the opposite of a sense entry in the tag table, which is the tagger's sentence about how the paper uses a term. Where a concept entry could not be grounded in the paper's words, the entry says so on its own row.

Other fields define these differently the same word, defined another way in another discipline
deep learning 11 fields · sample-size 5 fields · DFT 4 fields · graph neural network 3 fields · supervised learning 3 fields · graph 2 fields · and 2 more →

Tags on this paper

Every tag the record carries, each linking to every other paper in the corpus that carries it. A tag appears here whether or not its words occur in the article — the spans below are only the ones a sentence in this paper licenses. Of the 10 concept tags on this paper, 5 are grounded in a quoted sentence.

graph neural network message passing graph molecular property prediction deep learning chemical accuracy DFT error Message Passing Neural Networks ground truth labels QM9 message passing on the molecular graph (atoms as nodes, bonds as edges): iteratively passing and aggregating messages between atoms quantum chemistry molecular property graph isomorphism MPNN DFT readout function node features hidden states supervised learning Message Passing Neural Networks (MPNNs) readout phase time steps edge features update functions readout functions mean average of sample-size system-or-tool-building

Legend

relation to other work	`claim_relations`	how this paper stands to another: agrees, contradicts, extends, qualifies
claim type	`claim_type`	the shape of the claim: existence, mechanism, method, critique, prediction, null
concept edge	`concept_edges`	an asserted relation between two concepts, with the sentence asserting it
concept	`concepts`	an idea the paper works with, quoted where the paper uses it
content tag	`content_tags`	a finding, definition, quantity, statistic or variable named in the paper
control	`controls`	an experiment run to eliminate a rival explanation
limitation	`limitations`	what the paper says its evidence cannot support
mechanism	`mechanisms`	how the reported effect is said to come about
method	`method`	how the paper produced its evidence
phenomenon	`phenomenon`	the effect or regularity under study
sense	`senses`	a distinct use of a term, quoted where the paper fixes it
transfer	`transfers`	a method or concept carried in from another field

The paper

The full text is not reproduced here. This article is not cleared for redistribution, so the page carries the tag layer only. Its evidence sentences are withheld; the grounding status remains with each tag, and the article itself is at https://arxiv.org/abs/1704.01212. The reading view exists for the papers whose licence permits it.

Every tag on this page

claim_relations 4 · claim_type 1 · concept_edges 2 · concepts 5 · content_tags 20 · controls 1 · limitations 9 · mechanisms 1 · method 1 · phenomenon 1 · senses 3 · transfers 1

Relations on this page carry a quotation for the claim. Where a relation was separately evidenced — and so tested against the vocabulary for its type — the row says so; none of 2 relations here carry that mark. The rest are the tagger’s reading of the sentence quoted beside them.

facet	tag	field	conf.	whose words
claim_type	method Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`claim_type`		the paper's words
phenomenon	molecular property prediction from graphs Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`phenomenon`		the paper's words
method	system-or-tool-building Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: system-or-tool-building	`method`		the paper's words
mechanisms	mechanisms Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: message passing on the molecular graph (atoms as nodes, bonds as edges): iteratively passing and aggregating messages between atoms	`mechanisms[0]`		the tagger's words
limitations	the models fit the DFT approximation, itself an approximation of true quantum properties Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: every accuracy claim: the target is THE DFT APPROXIMATION, so 'chemical accuracy' is accuracy to a model rather than to the quantum properties themselves	`limitations[0]`		the tagger's words
limitations	The towers variant improves scalability but is still insufficient for much larger graphs. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: the scalability claim: THE TOWERS VARIATION MAKES THESE MODELS MORE SCALABLE, BUT ADDITIONAL IMPROVEMENTS WILL BE NEEDED TO SCALE TO MUCH LARGER GRAPHS	`limitations[1]`		the tagger's words
limitations	Generalising to larger molecules is hard precisely when spatial information is used. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: the spatial-information variants specifically: GENERALIZING TO LARGER MOLECULE SIZES SEEMS PARTICULARLY CHALLENGING WHEN USING SPATIAL INFORMATION	`limitations[2]`		the tagger's words
limitations	Combining the towers variant with the edge-network message function did not work. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: the towers-plus-edge-network combination: UNFORTUNATELY, OUR ATTEMPTS SO FAR AT COMBINING THE TOWERS AND EDGE NETWORK MESSAGE FUNCTION HAVE FAILED TO FURTHER IMPROVE PERFORMANCE	`limitations[3]`		the tagger's words
limitations	Some ablation experiments used a partial-charge input feature that would not be available in a real application. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: the ablations using partial charge, which no deployment could reproduce: THIS FEATURE IS AN OUTPUT OF THE DFT CALCULATION AND THUS COULD NOT BE USED IN AN APPLIED SETTING	`limitations[4]`		the tagger's words
limitations	Accuracy is measured against DFT approximations because experimentally true targets do not exist for the full dataset. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: every accuracy figure on this benchmark, which is agreement with a computation rather than with measurement: THE DATASET CONTAINING THE TRUE TARGETS ON ALL 134k MOLECULES DOES NOT CURRENTLY EXIST	`limitations[5]`		the tagger's words
limitations	The paper's headline term is redefined to mean error against its own labels rather than against nature, and it says so in one sentence. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: the claim that this model reaches chemical accuracy: the authors state that in the rest of the paper the phrase means accuracy with respect to their available ground truth labels, not the community's target error relative to nature	`limitations[6]`	1.0	the tagger's words
limitations	Two different error standards are in play and the paper orders them. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: a comparison of the errors reported here against either standard: the paper states that Quotation withheld	`limitations[7]`	1.0	the tagger's words
limitations	Eight existing models are re-described in the authors' own framework, with the derivations placed outside the paper. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. what this limits: the claim that this framework unifies prior work: at least eight models from the literature are described in it, and the exact expressions and the derivation of that reformulation are in the supplementary material rather than the paper	`limitations[8]`	1.0	the tagger's words
transfers	graph neural networks / deep learning, applied to predict quantum-chemical properties of molecules Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`transfers[0]`		the tagger's words
controls	the model family evaluated on all 13 targets of the public QM9 benchmark Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. eliminates: a chemistry result earned on molecules of the authors' own choosing - QM9 is fixed, public and shared with the methods being compared against · outcome (stated by the tagger, no quotation): held	`controls[0]`	0.85	the tagger's words
concepts	chemical accuracy Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: chemical accuracy	`concepts[5]`		the paper's words
concepts	DFT error Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: DFT error	`concepts[6]`		the paper's words
concepts	Message Passing Neural Networks Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: Message Passing Neural Networks	`concepts[7]`		the paper's words
concepts	ground truth labels Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: ground truth labels	`concepts[8]`		the paper's words
concepts	QM9 Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: QM9	`concepts[9]`		the paper's words
senses	'chemical accuracy' - TWO REFERENCE POINTS, AND THE PAPER PICKS THE NEARER ONE: the chemistry community's target error is relative to nature, DFT error is the gap between the simulation and nature, and this paper states that its own use of the phrase means error against its available labels - so the headline measures agreement with a simulation, not with a molecule Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[0]`	1.0	the tagger's words
senses	MPNN - A NAME GIVEN TO EIGHT EXISTING MODELS AFTER THE FACT: the framework is introduced as a reformulation of models already in the literature under one label, which is the inverse of the usual naming failure - here a paper renames its field rather than being renamed by it Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[1]`	1.0	the tagger's words
senses	the prediction target - A MODEL OF A MODEL: what the network predicts is the outcome of a computationally expensive DFT calculation, so the object being learned is itself a simulation of the quantity of interest rather than the quantity Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[2]`	1.0	the tagger's words
claim_relations	builds-on → graph neural networks (invariant to molecular symmetries) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`claim_relations[0]`		the tagger's words
claim_relations	builds-on → Gated Graph Neural Networks (GG-NN), Li et al. (2016) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`claim_relations[1]`		the tagger's words
claim_relations	builds-on → the set2set model, Vinyals et al. (2015) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`claim_relations[2]`		the tagger's words
claim_relations	extends → Interaction Networks, Battaglia et al. (2016) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`claim_relations[3]`		the tagger's words
concept_edges	chemical accuracy —operationalised-as→ ground truth labels Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. strength of the relation: 1 also in: chemical accuracy · ground truth labels	`concept_edges[0]`		the tagger's words
concept_edges	chemical accuracy —distinct-from→ DFT error Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. strength of the relation: 1 also in: chemical accuracy · DFT error	`concept_edges[1]`		the tagger's words
content_tags	130k — size of the QM9 benchmark with 13 properties each Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[0]`		the tagger's words
content_tags	300k — MPNN inference speedup over DFT simulation Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[1]`		the tagger's words
content_tags	11 — targets predicted within chemical accuracy by the best MPNN Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[2]`		the tagger's words
content_tags	5 — targets hit using molecular topology alone Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. about this quote: marked quote-not-in-current-staged-text - today's merge gate would REFUSE this quote: Paper.locate() does not find it in the current staged text, and no disclosure on the row explains why. Update, 15 September 2026: CLEARED evidence status='quote-not-in-current-staged-text' - the quote was replaced (EXTEND) and the new quote locates exactly once in the current staged text.	`content_tags.quantities[3]`		the tagger's words
content_tags	8 — prior models unified under the MPNN framework Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[4]`		the tagger's words
content_tags	MPNN variants achieve a new state of the art on the QM9 dataset Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.findings[0]`		the tagger's words
content_tags	neural inference is far cheaper than the DFT simulation it approximates Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.findings[1]`		the tagger's words
content_tags	the readout must be permutation-invariant for graph-isomorphism invariance Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.findings[2]`		the tagger's words
content_tags	Duvenaud-style message passing cannot capture edge-node correlations Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.findings[3]`		the tagger's words
content_tags	hidden states Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: hidden states	`content_tags.variables[1]`		the tagger's words
content_tags	time steps Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: time steps	`content_tags.variables[2]`		the tagger's words
content_tags	node features Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: node features	`content_tags.variables[4]`		the tagger's words
content_tags	edge features Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: edge features	`content_tags.variables[5]`		the tagger's words
content_tags	update functions Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: update functions	`content_tags.variables[6]`		the tagger's words
content_tags	readout functions Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: readout functions	`content_tags.variables[7]`		the tagger's words
content_tags	Message Passing Neural Networks (MPNNs) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: Message Passing Neural Networks (MPNNs)	`content_tags.definitions[0]`		the tagger's words
content_tags	chemical accuracy Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: chemical accuracy	`content_tags.definitions[1]`		the tagger's words
content_tags	DFT error Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: DFT error	`content_tags.definitions[2]`		the tagger's words
content_tags	readout phase Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: readout phase	`content_tags.definitions[3]`		the tagger's words
content_tags	sample-size Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: sample-size	`content_tags.statistics[1]`		the tagger's words

Interpretive choices

Tags that could not be placed in the text:

content_tags.statistics[0] — quotation withheld — the page has already reproduced as much of this paper as the rights budget allows
NOT MACHINE-VOUCHED: not-in-our-copy - The quote is in the original web page we fetched for this paper, but not in the text we tag against, which was extracted differently. Checked with ligatures, hyphens, spacing and punctuation set aside.

What was deliberately not tagged, and why, is recorded per paper by the tagger. Where a paper has no such record, this page says so rather than implying that nothing was declined.