Scaling deep learning for materials discovery

Amil Merchant, Simon Batzner, Samuel S. Schoenholz, Muratahan Aykol, Gowoon Cheon, Ekin D. Cubuk · Nature · 2023 · materials-science · 10.1038/s41586-023-06735-9 · record merchant2023-scaling-deep-learning-for-material
tag layer · index · concepts · authors · עברית

Where this record's text came from. 2026-08-20 and 2026-08-28 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.

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 WITHDRAWN AND REPLACED — the entry was wrong, and a corrected version of it appears in the entries below.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 An entry we have WITHDRAWN is removed from this page and kept in the record, with its reason.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
DFT 4 fields · active learning 3 fields

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 85 concept tags on this paper, all 85 are grounded in a quoted sentence. All carry the tagger's gloss of what the term means in this paper, which is not something the article can license.

convex hull decomposition energy hit rate active learning GNoME Materials Project r2SCAN experimental match prototypes graph networks trained at scale efficiency of materials discovery human chemical intuition learned interatomic potentials condensed-phase molecular-dynamics simulations zero-shot prediction of ionic conductivity large-scale active learning first models that accurately predict stability diverse candidate structures symmetry-aware partial substitutions (SAPS) random structure search graph neural networks (GNNs) material properties candidate structures filtered candidates DFT data flywheel more robust models larger datasets model-based filtration and DFT predictions structural candidates ionic substitution probabilities incomplete replacements volume-based test-time augmentation uncertainty quantification through deep ensembles polymorphs compositional models reduced chemical formulas oxidation-state balancing relaxed constraints ab initio random structure searching (AIRSS) relative stability (decomposition energy) Vienna Ab initio Simulation Package (VASP) number of stable materials discovered total energy of a crystal one-hot embedding of the elements message-passing formulation shallow multilayer perceptrons (MLPs) swish nonlinearities average adjacency of atoms mean absolute error (MAE) test loss performance power law with further data neural scaling laws out-of-distribution tasks high-energy local minima universal energy predictor ionic relaxations first-principles energies and forces general-purpose machine-learning interatomic potential (MLIP) bulk solids ab initio molecular dynamics (AIMD) general applicability and adoption large-scale pretraining of MLIPs unprecedented zero-shot accuracy superionic conductors material-specific data increasing the pretraining dataset power-law improvements in accuracy pretrained GNoME potential unseen compositions melted structures structures including vacancies transferability of MLIPs potential pretrained on GNoME data potential trained from scratch GNoME-pretrained randomly initialized potential distribution shift stable crystals identically distributed data at train and test times out-of-distribution effort emergent out-of-distribution capabilities unseen chemical spaces kinetic properties the data flywheel power-law scaling of prediction error power-law improvement of test loss with data precision of stable predictions, reported as a hit rate system-or-tool-building

Method, as described: rounds of active learning: generate candidates, filter with graph networks, verify the survivors with DFT, retrain on the result

Legend

second reading	`alternatives`	an independent re-reading of a facet, recorded whether or not it agrees
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
contribution	`contribution`	what the paper claims to add
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
object of study	`object_of_study`	what the paper is about — the thing it investigates
phenomenon	`phenomenon`	the effect or regularity under study
research question	`research_question`	the question the paper sets itself
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
unrecordable relation	`unhoused_relations`	a relation the paper asserts that this vocabulary cannot express, with the reason

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://doi.org/10.1038/s41586-023-06735-9. The reading view exists for the papers whose licence permits it.

Every tag on this page

alternatives 3 · claim_relations 2 · claim_type 1 · concept_edges 9 · concepts 85 · content_tags 20 · contribution 1 · controls 5 · limitations 4 · mechanisms 2 · method 1 · object_of_study 1 · phenomenon 1 · research_question 1 · senses 13 · transfers 1 · unhoused_relations 4

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; 7 of 9 relations here carry that mark. The rest are the tagger’s reading of the sentence quoted beside them.

facet	tag	field	conf.	whose words
object_of_study	inorganic crystal structures and whether a graph neural network can predict which are thermodynamically stable Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`object_of_study`		the paper's words
research_question	whether machine learning can estimate stability against the convex hull accurately enough to guide discovery, which prior methods could not Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`research_question`		the paper's words
contribution	2.2 million structures stable with respect to the Materials Project, of which 381,000 sit on the updated convex hull, plus learned interatomic potentials from the resulting dataset Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`contribution`		the paper's words
claim_type	existence 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	emergent out-of-distribution generalisation in a graph network trained at scale 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	DFT verifies the model's filtered candidates and the verified results become the next round's training data, so the same computation both audits the model and improves it Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: the data flywheel	`mechanisms[0]`	0.95	the tagger's words
mechanisms	prediction quality improves as a power law in the amount of data, matching what has been observed in language and vision rather than being specific to materials Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: power-law scaling of prediction error	`mechanisms[1]`	0.95	the tagger's words
limitations	the discovered materials can themselves be bumped off the convex hull by future work - stability is a claim about a catalogue at a moment Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`limitations[0]`	0.95	the tagger's words
limitations	the experimental-match count is a lower bound, because the compositional match is enforced strictly Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`limitations[1]`	0.95	the tagger's words
limitations	the matching answer moves by about 100 structures depending on parser and matcher tolerance settings Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`limitations[2]`	0.95	the tagger's words
limitations	the discovery functional was chosen for a speed-accuracy trade-off rather than accuracy, and 84% rather than 100% survive the more accurate one Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`limitations[3]`	0.95	the tagger's words
transfers	neural scaling laws, carried from language and vision into materials-stability prediction Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`transfers[0]`	0.95	the tagger's words
controls	recomputation under the more accurate r2SCAN functional, to test whether discovery overfitted the cheaper PBE one Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. eliminates: a discovery set that exists only under the cheap functional it was searched with · outcome: partial	`controls[0]`	0.95	the tagger's words
controls	an explicit check that model-based exploration did not overfit the simulation parameters Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. eliminates: model-based exploration finding materials that are artefacts of the simulation parameters it was trained against · outcome: held	`controls[1]`	0.95	the tagger's words
controls	matching against crystals experimentally realised by other groups after the training snapshot Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. eliminates: a match rate inflated by structures the model could have seen, since these were realised experimentally after the training snapshot · outcome: held	`controls[2]`	0.95	the tagger's words
controls	comparison against compositions added to the Materials Project after the March 2021 snapshot Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. eliminates: a match rate computed only against the database the model was trained on · outcome: held	`controls[3]`	0.95	the tagger's words
controls	a sensitivity check on the parser and matcher tolerances used to count experimental matches Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. eliminates: an experimental match count that exists only at one parser and matcher tolerance setting · outcome: partial	`controls[4]`	0.95	the tagger's words
concepts	convex hull Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: convex hull	`concepts[0]`	0.95	the paper's words
concepts	decomposition energy Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: decomposition energy	`concepts[1]`	0.95	the paper's words
concepts	hit rate Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: hit rate	`concepts[2]`	0.95	the paper's words
concepts	active learning Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: active learning	`concepts[3]`	0.95	the paper's words
concepts	GNoME Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: GNoME	`concepts[4]`	0.95	the paper's words
concepts	Materials Project Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: Materials Project	`concepts[5]`	0.95	the paper's words
concepts	r2SCAN Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: r2SCAN	`concepts[6]`	0.95	the paper's words
concepts	experimental match Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: experimental match	`concepts[7]`	0.95	the paper's words
concepts	prototypes Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: prototypes	`concepts[8]`	0.95	the paper's words
concepts	graph networks trained at scale Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: graph networks trained at scale	`concepts[9]`	0.95	the paper's words
concepts	efficiency of materials discovery Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: efficiency of materials discovery	`concepts[10]`	0.95	the paper's words
concepts	human chemical intuition Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: human chemical intuition	`concepts[11]`	0.95	the paper's words
concepts	learned interatomic potentials Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: learned interatomic potentials	`concepts[12]`	0.95	the paper's words
concepts	condensed-phase molecular-dynamics simulations Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: condensed-phase molecular-dynamics simulations	`concepts[13]`	0.95	the paper's words
concepts	zero-shot prediction of ionic conductivity Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: zero-shot prediction of ionic conductivity	`concepts[14]`	0.95	the paper's words
concepts	large-scale active learning Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: large-scale active learning	`concepts[15]`	0.95	the paper's words
concepts	first models that accurately predict stability Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: first models that accurately predict stability	`concepts[16]`	0.95	the paper's words
concepts	diverse candidate structures Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: diverse candidate structures	`concepts[17]`	0.95	the paper's words
concepts	symmetry-aware partial substitutions (SAPS) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: symmetry-aware partial substitutions (SAPS)	`concepts[18]`	0.95	the paper's words
concepts	random structure search Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: random structure search	`concepts[19]`	0.95	the paper's words
concepts	graph neural networks (GNNs) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: graph neural networks (GNNs)	`concepts[20]`	0.95	the paper's words
concepts	material properties Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: material properties	`concepts[21]`	0.95	the paper's words
concepts	candidate structures Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: candidate structures	`concepts[22]`	0.95	the paper's words
concepts	filtered candidates Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: filtered candidates	`concepts[23]`	0.95	the paper's words
concepts	DFT Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: DFT	`concepts[24]`	0.95	the paper's words
concepts	data flywheel Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: data flywheel	`concepts[25]`	0.95	the paper's words
concepts	more robust models Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: more robust models	`concepts[26]`	0.95	the paper's words
concepts	larger datasets Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: larger datasets	`concepts[27]`	0.95	the paper's words
concepts	model-based filtration and DFT Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: model-based filtration and DFT	`concepts[28]`	0.95	the paper's words
concepts	predictions Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: predictions	`concepts[29]`	0.95	the paper's words
concepts	structural candidates Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: structural candidates	`concepts[30]`	0.95	the paper's words
concepts	ionic substitution probabilities Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: ionic substitution probabilities	`concepts[31]`	0.95	the paper's words
concepts	incomplete replacements Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: incomplete replacements	`concepts[32]`	0.95	the paper's words
concepts	volume-based test-time augmentation Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: volume-based test-time augmentation	`concepts[33]`	0.95	the paper's words
concepts	uncertainty quantification through deep ensembles Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: uncertainty quantification through deep ensembles	`concepts[34]`	0.95	the paper's words
concepts	polymorphs Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: polymorphs	`concepts[35]`	0.95	the paper's words
concepts	compositional models Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: compositional models	`concepts[36]`	0.95	the paper's words
concepts	reduced chemical formulas Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: reduced chemical formulas	`concepts[37]`	0.95	the paper's words
concepts	oxidation-state balancing Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: oxidation-state balancing	`concepts[38]`	0.95	the paper's words
concepts	relaxed constraints Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: relaxed constraints	`concepts[39]`	0.95	the paper's words
concepts	ab initio random structure searching (AIRSS) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: ab initio random structure searching (AIRSS)	`concepts[40]`	0.95	the paper's words
concepts	relative stability (decomposition energy) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: relative stability (decomposition energy)	`concepts[41]`	0.95	the paper's words
concepts	Vienna Ab initio Simulation Package (VASP) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: Vienna Ab initio Simulation Package (VASP)	`concepts[42]`	0.95	the paper's words
concepts	number of stable materials discovered Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: number of stable materials discovered	`concepts[43]`	0.95	the paper's words
concepts	total energy of a crystal Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: total energy of a crystal	`concepts[44]`	0.95	the paper's words
concepts	one-hot embedding of the elements Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: one-hot embedding of the elements	`concepts[45]`	0.95	the paper's words
concepts	message-passing formulation Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: message-passing formulation	`concepts[46]`	0.95	the paper's words
concepts	shallow multilayer perceptrons (MLPs) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: shallow multilayer perceptrons (MLPs)	`concepts[47]`	0.95	the paper's words
concepts	swish nonlinearities Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: swish nonlinearities	`concepts[48]`	0.95	the paper's words
concepts	average adjacency of atoms Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: average adjacency of atoms	`concepts[49]`	0.95	the paper's words
concepts	mean absolute error (MAE) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: mean absolute error (MAE)	`concepts[50]`	0.95	the paper's words
concepts	test loss performance Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: test loss performance	`concepts[51]`	0.95	the paper's words
concepts	power law with further data Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: power law with further data	`concepts[52]`	0.95	the paper's words
concepts	neural scaling laws Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: neural scaling laws	`concepts[53]`	0.95	the paper's words
concepts	out-of-distribution tasks Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: out-of-distribution tasks	`concepts[54]`	0.95	the paper's words
concepts	high-energy local minima Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: high-energy local minima	`concepts[55]`	0.95	the paper's words
concepts	universal energy predictor Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: universal energy predictor	`concepts[56]`	0.95	the paper's words
concepts	ionic relaxations Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: ionic relaxations	`concepts[57]`	0.95	the paper's words
concepts	first-principles energies and forces Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: first-principles energies and forces	`concepts[58]`	0.95	the paper's words
concepts	general-purpose machine-learning interatomic potential (MLIP) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: general-purpose machine-learning interatomic potential (MLIP)	`concepts[59]`	0.95	the paper's words
concepts	bulk solids Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: bulk solids	`concepts[60]`	0.95	the paper's words
concepts	ab initio molecular dynamics (AIMD) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: ab initio molecular dynamics (AIMD)	`concepts[61]`	0.95	the paper's words
concepts	general applicability and adoption Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: general applicability and adoption	`concepts[62]`	0.95	the paper's words
concepts	large-scale pretraining of MLIPs Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: large-scale pretraining of MLIPs	`concepts[63]`	0.95	the paper's words
concepts	unprecedented zero-shot accuracy Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: unprecedented zero-shot accuracy	`concepts[64]`	0.95	the paper's words
concepts	superionic conductors Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: superionic conductors	`concepts[65]`	0.95	the paper's words
concepts	material-specific data Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: material-specific data	`concepts[66]`	0.95	the paper's words
concepts	increasing the pretraining dataset Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: increasing the pretraining dataset	`concepts[67]`	0.95	the paper's words
concepts	power-law improvements in accuracy Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: power-law improvements in accuracy	`concepts[68]`	0.95	the paper's words
concepts	pretrained GNoME potential Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: pretrained GNoME potential	`concepts[69]`	0.95	the paper's words
concepts	unseen compositions Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: unseen compositions	`concepts[70]`	0.95	the paper's words
concepts	melted structures Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: melted structures	`concepts[71]`	0.95	the paper's words
concepts	structures including vacancies Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: structures including vacancies	`concepts[72]`	0.95	the paper's words
concepts	transferability of MLIPs Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: transferability of MLIPs	`concepts[73]`	0.95	the paper's words
concepts	potential pretrained on GNoME data Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: potential pretrained on GNoME data	`concepts[74]`	0.95	the paper's words
concepts	potential trained from scratch Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: potential trained from scratch	`concepts[75]`	0.95	the paper's words
concepts	GNoME-pretrained Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: GNoME-pretrained	`concepts[76]`	0.95	the paper's words
concepts	randomly initialized potential Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: randomly initialized potential	`concepts[77]`	0.95	the paper's words
concepts	distribution shift Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: distribution shift	`concepts[78]`	0.95	the paper's words
concepts	stable crystals Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: stable crystals	`concepts[79]`	0.95	the paper's words
concepts	identically distributed data at train and test times Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: identically distributed data at train and test times	`concepts[80]`	0.95	the paper's words
concepts	out-of-distribution effort Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: out-of-distribution effort	`concepts[81]`	0.95	the paper's words
concepts	emergent out-of-distribution capabilities Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: emergent out-of-distribution capabilities	`concepts[82]`	0.95	the paper's words
concepts	unseen chemical spaces Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: unseen chemical spaces	`concepts[83]`	0.95	the paper's words
concepts	kinetic properties Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: kinetic properties	`concepts[84]`	0.95	the paper's words
unhoused_relations	graph neural networks (GNNs) → candidate structures Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. GNoME graph networks are used to filter candidate structures. Used-to-filter is absent from the eight types.	`unhoused_relations[0]`	0.95	the tagger's words
unhoused_relations	DFT → model predictions Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. DFT verifies the model predictions. Verifies is absent from the eight types.	`unhoused_relations[1]`	0.95	the tagger's words
unhoused_relations	GNoME dataset → general-purpose machine-learning interatomic potential (MLIP) Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. The GNoME dataset enables pretraining a general-purpose MLIP. Enablement is absent from the eight types.	`unhoused_relations[2]`	0.95	the tagger's words
unhoused_relations	final GNoME models → universal energy predictor Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. Final GNoME models are a substantial step toward a universal energy predictor. Progress-toward is absent from the eight types.	`unhoused_relations[3]`	0.95	the tagger's words
senses	'stable' sense 1 = below the convex hull OF A PRIOR CATALOGUE - 2.2 million structures Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[0]`	0.95	the tagger's words
senses	'stable' sense 2 = ON the UPDATED convex hull after this work - 381,000 entries, six times fewer Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[1]`	0.95	the tagger's words
senses	stability is REVOCABLE: these materials can be bumped off the hull by future discoveries, exactly as this work bumps off 5,000 others Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[2]`	0.95	the tagger's words
senses	'meaningfully stable' = distributed like known stable phases, not merely filling gaps in the hull Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[3]`	0.95	the tagger's words
senses	'experimental match' = integer-rounded composition match AND a pymatgen structure match, declared a lower bound Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[4]`	0.95	the tagger's words
senses	GNoME discovery flywheel - generate candidates, filter with GNNs, verify energies with DFT, and reuse results as next-round training data Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[5]`	0.95	the tagger's words
senses	Structural framework - modify known crystals, expand substitutions with SAPS, quantify uncertainty, cluster polymorphs, and rank for DFT Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[6]`	0.95	the tagger's words
senses	Compositional framework - predict from reduced formulas, relax oxidation constraints, filter compositions, and initialize AIRSS structures Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[7]`	0.95	the tagger's words
senses	GNoME graph model - element embeddings and message passing predict total crystal energy Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[8]`	0.95	the tagger's words
senses	Scaling law - GNoME test loss improves by a power law as training data increase Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[9]`	0.95	the tagger's words
senses	Universal energy predictor - model handling diverse material structures including random-search inputs outside the training distribution Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[10]`	0.95	the tagger's words
senses	General-purpose MLIP - potential pretrained on diverse ionic-relaxation energies and forces for zero-shot bulk-solid use Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[11]`	0.95	the tagger's words
senses	Zero-shot evaluation - pretrained potential applied to unseen compositions, melted structures, vacancies, and AIMD data without AIMD training Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`senses[12]`	0.95	the tagger's words
alternatives	claim_type: method Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. a second reading DIFFERING from the recorded value. Recorded: existence — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is).,	`alternatives.claim_type`	0.7	the tagger's words
alternatives	method: system-or-tool-building Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. a second reading DIFFERING from the recorded value. Recorded: system-or-tool-building — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is).,	`alternatives.method`	0.8	the tagger's words
alternatives	phenomenon: discovery of energetically stable inorganic crystal structures at a scale experiment cannot reach Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. a second reading DIFFERING from the recorded value. Recorded: emergent out-of-distribution generalisation in a graph network trained at scale — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is).,	`alternatives.phenomenon`	0.85	the tagger's words
claim_relations	contradicts → the prior position that machine learning cannot estimate stability against the convex hull Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. note: the paper states that position as the state of the art it overturns, and the hit-rate figures are what it offers against it	`claim_relations[0]`		the tagger's words
claim_relations	extends → the Materials Project, OQMD, AFLOWLIB and NOMAD catalogues of computationally stable structures Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. note: extends by an order of magnitude AND revises downward - at least 5,000 of their entries are displaced	`claim_relations[1]`		the tagger's words
concept_edges	GNoME —contrasted-with→ Materials Project Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. differs in: scale, and the direction of revision - GNoME removes at least 5,000 entries the Materials Project called stable · strength of the relation: 0.95 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span says GNoME displaces some 'stable' materials from the Materials Project; contrasted-with is loose. also in: GNoME · Materials Project	`concept_edges[0]`		the tagger's words
concept_edges	active learning —causes→ hit rate Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. direction on the measure: increases · strength of the relation: 0.85 · direction on the measure (stated by the tagger, no quotation): increases also in: active learning · hit rate	`concept_edges[1]`		the tagger's words
concept_edges	graph networks trained at scale —causes→ efficiency of materials discovery Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. direction on the measure: increases · the relation itself is separately evidenced · strength of the relation: 1 · note: Scaled graph networks improve discovery efficiency by an order of magnitude. also in: graph networks trained at scale · efficiency of materials discovery	`concept_edges[3]`	0.95	the tagger's words
concept_edges	large-scale active learning —causes→ first models that accurately predict stability Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. direction on the measure: increases · the relation itself is separately evidenced · strength of the relation: 1 · note: The paper says large-scale active learning yields the first accurate stability models. also in: large-scale active learning · first models that accurately predict stability	`concept_edges[4]`	0.95	the tagger's words
concept_edges	model-based filtration and DFT —causes→ predictions Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. direction on the measure: increases · the relation itself is separately evidenced · strength of the relation: 1 · note: The combined filtration/DFT flywheel is stated to improve predictions. Audit 2026-09-28: DOUBTFUL - read against its span and kept, but filtration and DFT generate candidates; causes 'predictions' is loose. also in: model-based filtration and DFT · predictions	`concept_edges[5]`	0.95	the tagger's words
concept_edges	increasing the pretraining dataset —causes→ power-law improvements in accuracy Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. direction on the measure: increases · the relation itself is separately evidenced · strength of the relation: 1 · note: Increasing pretraining scale produces consistent power-law accuracy improvements. also in: increasing the pretraining dataset · power-law improvements in accuracy	`concept_edges[7]`	0.95	the tagger's words
concept_edges	GNoME-pretrained —contrasted-with→ randomly initialized potential Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. the relation itself is separately evidenced · strength of the relation: 1 · note: The figure compares pretrained and randomly initialized potentials under distribution shift. also in: GNoME-pretrained · randomly initialized potential	`concept_edges[8]`	0.95	the tagger's words
concept_edges	potential pretrained on GNoME data —causes→ transferability of MLIPs Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. direction on the measure: increases · the relation itself is separately evidenced · strength of the relation: 1 · note: The pretrained potential shows systematic, strong transferability improvements over training from scratch. also in: potential pretrained on GNoME data · transferability of MLIPs	`concept_edges[9]`	0.95	the tagger's words
concept_edges	graph neural networks (GNNs) —causes→ stable crystals Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. direction on the measure: increases · the relation itself is separately evidenced · strength of the relation: 1 · note: Large, diverse first-principles training increases stable-crystal discovery by more than an order of magnitude. Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span writes 'GNNs'; endpoint abbreviated. also in: graph neural networks (GNNs) · stable crystals	`concept_edges[10]`	0.95	the tagger's words
content_tags	GNoME found 2.2 million structures stable relative to the Materials Project, of which 381,000 lie on the updated convex hull. Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.findings[0]`	0.95	the tagger's words
content_tags	this work removes at least 5,000 entries that two prior catalogues had called stable Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.findings[1]`	0.95	the tagger's words
content_tags	the networks generalise to structures with 5+ unique elements although such structures were omitted from training Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.findings[2]`	0.95	the tagger's words
content_tags	the discovered materials are distributed like known stable phases rather than merely filling gaps in the hull Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.findings[3]`	0.95	the tagger's words
content_tags	convex hull Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: convex hull	`content_tags.definitions[0]`	0.95	the tagger's words
content_tags	hit rate Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: hit rate	`content_tags.definitions[1]`	0.95	the tagger's words
content_tags	experimental match Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: experimental match	`content_tags.definitions[2]`	0.95	the tagger's words
content_tags	2.2 million structures stable with respect to the Materials Project; 381,000 on the updated hull — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[0]`	0.95	the tagger's words
content_tags	at least 5,000 previously 'stable' materials displaced from the Materials Project and OQMD — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[1]`	0.95	the tagger's words
content_tags	energies predicted to 11 meV per atom; hit rate above 80% with structure, 33% per 100 trials with composition, against 1% before — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[2]`	0.95	the tagger's words
content_tags	736 of the ICSD's experimental structures match GNoME discoveries; 184 are novel since the project began — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[3]`	0.95	the tagger's words
content_tags	4,235 composition matches narrowed to 4,180 parsed and then to 736 — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[4]`	0.95	the tagger's words
content_tags	84% of binaries and ternaries and 86.8% of quaternaries survive r2SCAN — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[5]`	0.95	the tagger's words
content_tags	45,500 novel prototypes, a 5.6 times increase on the Materials Project's 8,000 — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[6]`	0.95	the tagger's words
content_tags	layered materials rise from about 1,000 to about 52,000 — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[7]`	0.95	the tagger's words
content_tags	the ICSD holds 20,000 computationally stable structures out of 200,000 entries — None Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above.	`content_tags.quantities[8]`	0.95	the tagger's words
content_tags	power-law improvement of test loss with data Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: power-law improvement of test loss with data	`content_tags.statistics[0]`	0.95	the tagger's words
content_tags	precision of stable predictions, reported as a hit rate Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: precision of stable predictions, reported as a hit rate	`content_tags.statistics[1]`	0.95	the tagger's words
content_tags	decomposition energy Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: decomposition energy	`content_tags.variables[0]`	0.95	the tagger's words
content_tags	hit rate Quotation withheld — this record is not cleared for redistribution. The sentence is in the paper at the link above. also in: hit rate	`content_tags.variables[1]`	0.95	the tagger's words

Interpretive choices

Tags that could not be placed in the text:

none — every tag was located in the text

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.