phenomenon · claim-type · cross-domain-transfer · mechanism · claim-relation · replication-status · limitation — grounded in the full text| phenomenon | emergent out-of-distribution generalisation in a graph network trained at scale— “Moreover, these networks develop emergent out-of-distribution generalization. For example, GNoME enables accurate predictions of structures with 5+ unique elements (despite omission from training), providing one of the first strategies to efficiently explore this chemical space.” | |
| claim-type | existence— “In particular, GNoME models found 2.2 million crystal structures stable with respect to the Materials Project. Of these, 381,000 entries live on the updated convex hull as newly discovered materials.” | |
| method — evidence | — “The energy of the filtered candidates is computed using DFT, both verifying model predictions and serving as a data flywheel to train more robust models on larger datasets in the next round of active learning.” | |
| object of study — evidence | — “In particular, GNoME models found 2.2 million crystal structures stable with respect to the Materials Project. Of these, 381,000 entries live on the updated convex hull as newly discovered materials.” | |
| research question — evidence | — “Although data-driven methods that aid in further materials discovery have been pursued, thus far, machine-learning techniques have been ineffective in estimating stability (decomposition energy) with respect to the convex hull of energies from competing phases.” | |
| contribution — evidence | — “In particular, GNoME models found 2.2 million crystal structures stable with respect to the Materials Project. Of these, 381,000 entries live on the updated convex hull as newly discovered materials.” | |
| transfer | from machine learning: neural scaling laws, carried from language and vision into materials-stability prediction— “Consistent with observations in other domains of machine learning, we observe that our neural networks predictions improve as a power law with the amount of data.” | transfers[0] |
| mechanism | the data flywheel 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— “The energy of the filtered candidates is computed using DFT, both verifying model predictions and serving as a data flywheel to train more robust models on larger datasets in the next round of active learning.” | mechanisms[0] |
| mechanism | power-law scaling of prediction error 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— “Consistent with observations in other domains of machine learning, we observe that our neural networks predictions improve as a power law with the amount of data.” | mechanisms[1] |
| contradicts | the prior position that machine learning cannot estimate stability against the convex hull— “Although data-driven methods that aid in further materials discovery have been pursued, thus far, machine-learning techniques have been ineffective in estimating stability (decomposition energy) with respect to the convex hull of energies from competing phases.” | claim_relations[0] |
| extends | the Materials Project, OQMD, AFLOWLIB and NOMAD catalogues of computationally stable structures— “Combining ab initio calculations with simple substitutions has allowed researchers to improve to 48,000 computationally stable materials according to our own recalculations (see Methods).” | claim_relations[1] |
| sense | 'stable' sense 1 = below the convex hull OF A PRIOR CATALOGUE - 2.2 million structures— “In particular, GNoME models found 2.2 million crystal structures stable with respect to the Materials Project. Of these, 381,000 entries live on the updated convex hull as newly discovered materials.” | senses[0] |
| sense | 'stable' sense 2 = ON the UPDATED convex hull after this work - 381,000 entries, six times fewer— “In particular, GNoME models found 2.2 million crystal structures stable with respect to the Materials Project. Of these, 381,000 entries live on the updated convex hull as newly discovered materials.” | senses[1] |
| sense | stability is REVOCABLE: these materials can be bumped off the hull by future discoveries, exactly as this work bumps off 5,000 others— “Consistent with other literature on structure prediction, the GNoME materials could be bumped off the convex hull by future discoveries, similar to how GNoME displaces at least 5,000 ‘stable’ materials from the Materials Project and the OQMD.” | senses[2] |
| sense | 'meaningfully stable' = distributed like known stable phases, not merely filling gaps in the hull— “The similarities in distribution suggest that the found materials are meaningfully stable with respect to competing phases and not just ‘filling in the convex hull.’” | senses[3] |
| sense | 'experimental match' = integer-rounded composition match AND a pymatgen structure match, declared a lower bound— “Given that we are enforcing a strict compositional match, our matching process is still relatively conservative and is likely to yield a lower bound.” | senses[4] |
| sense | GNoME discovery flywheel - generate candidates, filter with GNNs, verify energies with DFT, and reuse results as next-round training data— “In this paper, we scale up machine learning for materials exploration through large-scale active learning, yielding the first models that accurately predict stability and, therefore, can guide materials discovery. Our approach relies on two pillars: first, we establish methods for generating diverse candidate structures, including new… | senses[5] |
| sense | Structural framework - modify known crystals, expand substitutions with SAPS, quantify uncertainty, cluster polymorphs, and rank for DFT— “To generate and filter candidates, we use two frameworks, which are visualized in Fig. 1a. First, structural candidates are generated by modifications of available crystals. However, we strongly augment the set of substitutions by adjusting ionic substitution probabilities to give priority to discovery and use newly proposed symmetry aware… | senses[6] |
| sense | Compositional framework - predict from reduced formulas, relax oxidation constraints, filter compositions, and initialize AIRSS structures— “To generate and filter candidates, we use two frameworks, which are visualized in Fig. 1a. First, structural candidates are generated by modifications of available crystals. However, we strongly augment the set of substitutions by adjusting ionic substitution probabilities to give priority to discovery and use newly proposed symmetry aware… | senses[7] |
| sense | GNoME graph model - element embeddings and message passing predict total crystal energy— “All GNoME models are GNNs that predict the total energy of a crystal. Inputs are converted to a graph through a one-hot embedding of the elements. We follow the message-passing formulation, in which aggregate projections are shallow multilayer perceptrons (MLPs) with swish nonlinearities. For structural models, we find it important to… | senses[8] |
| sense | Scaling law - GNoME test loss improves by a power law as training data increase— “The test loss performance of GNoME models exhibit improvement as a power law with further data. These results are in line with neural scaling laws in deep learning and suggest that further discovery efforts could continue to improve generalization. Emphatically, unlike the case of language or vision, in materials science, we can continue to… | senses[9] |
| sense | Universal energy predictor - model handling diverse material structures including random-search inputs outside the training distribution— “The test loss performance of GNoME models exhibit improvement as a power law with further data. These results are in line with neural scaling laws in deep learning and suggest that further discovery efforts could continue to improve generalization. Emphatically, unlike the case of language or vision, in materials science, we can continue to… | senses[10] |
| sense | General-purpose MLIP - potential pretrained on diverse ionic-relaxation energies and forces for zero-shot bulk-solid use— “The process of discovery of stable crystals also provides a data source beyond stable materials. In particular, the ionic relaxations involve computation of first-principles energies and forces for a diverse set of materials structures. This generates a dataset of unprecedented diversity and scale, which we explore to pretrain a… | senses[11] |
| sense | Zero-shot evaluation - pretrained potential applied to unseen compositions, melted structures, vacancies, and AIMD data without AIMD training— “We scale pretraining of a NequIP potential on data sampled from ionic relaxations. Increasing the pretraining dataset, we observe consistent power-law… | senses[12] |
| control | recomputation under the more accurate r2SCAN functional, to test whether discovery overfitted the cheaper PBE one — partial — rules out a discovery set that exists only under the cheap functional it was searched with Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[0] |
| control | an explicit check that model-based exploration did not overfit the simulation parameters — held Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. — rules out model-based exploration finding materials that are artefacts of the simulation parameters it was trained against Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[1] |
| control | matching against crystals experimentally realised by other groups after the training snapshot — held Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. — rules out a match rate inflated by structures the model could have seen, since these were realised experimentally after the training snapshot Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[2] |
| control | comparison against compositions added to the Materials Project after the March 2021 snapshot — held Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. — rules out a match rate computed only against the database the model was trained on Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[3] |
| control | a sensitivity check on the parser and matcher tolerances used to count experimental matches — partial Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. — rules out an experimental match count that exists only at one parser and matcher tolerance setting Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[4] |
| note | the prior position that machine learning cannot estimate stability against the convex hull — the paper states that position as the state of the art it overturns, and the hit-rate figures are what it offers against it Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | claim_relations[0] |
| weight | the prior position that machine learning cannot estimate stability against the convex hull — 0.95 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | claim_relations[0] |
| note | the Materials Project, OQMD, AFLOWLIB and NOMAD catalogues of computationally stable structures — extends by an order of magnitude AND revises downward - at least 5,000 of their entries are displaced Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | claim_relations[1] |
| weight | the Materials Project, OQMD, AFLOWLIB and NOMAD catalogues of computationally stable structures — 0.95 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | claim_relations[1] |
| audit | GNoME —contrasted-with→ Materials Project — 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. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[0] |
| differs in | GNoME —contrasted-with→ Materials Project — scale, and the direction of revision - GNoME removes at least 5,000 entries the Materials Project called stable Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[0] |
| weight | GNoME —contrasted-with→ Materials Project — 0.95 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[0] |
| effect direction (stated by the tagger, no quotation) | active learning —causes→ hit rate — increases | concept_edges[1] |
| weight | active learning —causes→ hit rate — 0.85 Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[1] |
| effect direction | graph networks trained at scale —causes→ efficiency of materials discovery — increases Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[3] |
| note | graph networks trained at scale —causes→ efficiency of materials discovery — Scaled graph networks improve discovery efficiency by an order of magnitude. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[3] |
| effect direction | large-scale active learning —causes→ first models that accurately predict stability — increases Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[4] |
| note | large-scale active learning —causes→ first models that accurately predict stability — The paper says large-scale active learning yields the first accurate stability models. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[4] |
| audit | model-based filtration and DFT —causes→ predictions — Audit 2026-09-28: DOUBTFUL - read against its span and kept, but filtration and DFT generate candidates; causes 'predictions' is loose. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[5] |
| effect direction | model-based filtration and DFT —causes→ predictions — increases Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[5] |
| note | model-based filtration and DFT —causes→ predictions — The combined filtration/DFT flywheel is stated to improve predictions. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[5] |
| effect direction | increasing the pretraining dataset —causes→ power-law improvements in accuracy — increases Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[7] |
| note | increasing the pretraining dataset —causes→ power-law improvements in accuracy — Increasing pretraining scale produces consistent power-law accuracy improvements. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[7] |
| note | GNoME-pretrained —contrasted-with→ randomly initialized potential — The figure compares pretrained and randomly initialized potentials under distribution shift. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[8] |
| effect direction | potential pretrained on GNoME data —causes→ transferability of MLIPs — increases Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[9] |
| note | potential pretrained on GNoME data —causes→ transferability of MLIPs — The pretrained potential shows systematic, strong transferability improvements over training from scratch. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[9] |
| audit | graph neural networks (GNNs) —causes→ stable crystals — Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span writes 'GNNs'; endpoint abbreviated. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[10] |
| effect direction | graph neural networks (GNNs) —causes→ stable crystals — increases Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[10] |
| note | graph neural networks (GNNs) —causes→ stable crystals — Large, diverse first-principles training increases stable-crystal discovery by more than an order of magnitude. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concept_edges[10] |
| outcome | control — partial Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[0] |
| outcome | control — held Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[1] |
| outcome | control — held Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[2] |
| outcome | control — held Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[3] |
| outcome | control — partial Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | controls[4] |
| no type for this | graph neural networks (GNNs) → candidate structures — GNoME graph networks are used to filter candidate structures. Used-to-filter is absent from the eight types. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[0] |
| no type for this | DFT → model predictions — DFT verifies the model predictions. Verifies is absent from the eight types. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[1] |
| no type for this | GNoME dataset → general-purpose machine-learning interatomic potential (MLIP) — The GNoME dataset enables pretraining a general-purpose MLIP. Enablement is absent from the eight types. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[2] |
| no type for this | final GNoME models → universal energy predictor — Final GNoME models are a substantial step toward a universal energy predictor. Progress-toward is absent from the eight types. Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | unhoused_relations[3] |
| limitation | 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 paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[0] |
| limitation | the experimental-match count is a lower bound, because the compositional match is enforced strictly Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[1] |
| limitation | the matching answer moves by about 100 structures depending on parser and matcher tolerance settings Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[2] |
| limitation | 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 paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | limitations[3] |
finding · quantity · variable · definition · key-term — grounded in the full text| finding | 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 paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[0] |
| finding | this work removes at least 5,000 entries that two prior catalogues had called stable Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[1] |
| finding | the networks generalise to structures with 5+ unique elements although such structures were omitted from training Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[2] |
| finding | the discovered materials are distributed like known stable phases rather than merely filling gaps in the hull Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.findings[3] |
| quantity | 2.2 million structures stable with respect to the Materials Project; 381,000 on the updated hull — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[0] |
| quantity | at least 5,000 previously 'stable' materials displaced from the Materials Project and OQMD — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[1] |
| quantity | energies predicted to 11 meV per atom; hit rate above 80% with structure, 33% per 100 trials with composition, against 1% before — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[2] |
| quantity | 736 of the ICSD's experimental structures match GNoME discoveries; 184 are novel since the project began — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[3] |
| quantity | 4,235 composition matches narrowed to 4,180 parsed and then to 736 — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[4] |
| quantity | 84% of binaries and ternaries and 86.8% of quaternaries survive r2SCAN — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[5] |
| quantity | 45,500 novel prototypes, a 5.6 times increase on the Materials Project's 8,000 — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[6] |
| quantity | layered materials rise from about 1,000 to about 52,000 — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[7] |
| quantity | the ICSD holds 20,000 computationally stable structures out of 200,000 entries — Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.quantities[8] |
| variable | decomposition energy Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.variables[0] |
| variable | hit rate Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.variables[1] |
| definition | convex hull Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.definitions[0] |
| definition | hit rate Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.definitions[1] |
| definition | experimental match Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.definitions[2] |
| concept | convex hull — not defined in this paper's own words; the surface of lowest-energy competing phases; being on it is what 'stable' means here Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[0] |
| concept | decomposition energy — not defined in this paper's own words; named in the paper as the thing 'stability' operationally is Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[1] |
| concept | hit rate — not defined in this paper's own words; the precision of stable predictions - above 80% with structure, 33% per 100 trials with composition Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[2] |
| concept | active learning — not defined in this paper's own words; rounds in which DFT both verifies the model and supplies its next training set Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[3] |
| concept | GNoME — not defined in this paper's own words; graph networks for materials exploration, the system this paper announces Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[4] |
| concept | Materials Project — not defined in this paper's own words; the catalogue stability is measured against - and the one this work displaces entries from Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[5] |
| concept | r2SCAN — not defined in this paper's own words; the more accurate functional used to check that discovery did not overfit the cheaper one Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[6] |
| concept | experimental match — not defined in this paper's own words; a composition match rounded to integer formulas, then a pymatgen structure match Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[7] |
| concept | prototypes — not defined in this paper's own words; structure types; 45,500 novel ones, a 5.6x increase, used as the diversity argument Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[8] |
| concept | graph networks trained at scale — not defined in this paper's own words; scaled models that generalize and accelerate materials discovery Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[9] |
| concept | efficiency of materials discovery — not defined in this paper's own words; discovery efficiency improved by an order of magnitude Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[10] |
| concept | human chemical intuition — not defined in this paper's own words; prior search guidance escaped by many predicted structures Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[11] |
| concept | learned interatomic potentials — not defined in this paper's own words; downstream models trained from first-principles trajectories Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[12] |
| concept | condensed-phase molecular-dynamics simulations — not defined in this paper's own words; downstream simulation use of learned potentials Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[13] |
| concept | zero-shot prediction of ionic conductivity — not defined in this paper's own words; downstream property prediction without task-specific training Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[14] |
| concept | large-scale active learning — not defined in this paper's own words; iterative scaling strategy for materials exploration Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[15] |
| concept | first models that accurately predict stability — not defined in this paper's own words; models yielded by large-scale active learning Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[16] |
| concept | diverse candidate structures — not defined in this paper's own words; candidate set generated by the first pipeline pillar Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[17] |
| concept | symmetry-aware partial substitutions (SAPS) — not defined in this paper's own words; candidate-generation method for incomplete substitutions Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[18] |
| concept | random structure search — not defined in this paper's own words; second diverse candidate-generation method Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[19] |
| concept | graph neural networks (GNNs) — not defined in this paper's own words; structure- or composition-conditioned property models Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[20] |
| concept | material properties — not defined in this paper's own words; quantities modelled from structure or composition Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[21] |
| concept | candidate structures — not defined in this paper's own words; generated structures filtered by GNoME Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[22] |
| concept | filtered candidates — not defined in this paper's own words; candidates evaluated by density-functional theory Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[23] |
| concept | DFT — not defined in this paper's own words; first-principles evaluator verifying model predictions Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[24] |
| concept | data flywheel — not defined in this paper's own words; DFT feedback that supplies the next active-learning round Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[25] |
| concept | more robust models — not defined in this paper's own words; models trained on larger feedback datasets Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[26] |
| concept | larger datasets — not defined in this paper's own words; data accumulated for later active-learning rounds Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[27] |
| concept | model-based filtration and DFT — not defined in this paper's own words; combined flywheel that improves predictions Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[28] |
| concept | predictions — not defined in this paper's own words; model outputs improved by the discovery flywheel Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[29] |
| concept | structural candidates — not defined in this paper's own words; crystal modifications used in the structural framework Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[30] |
| concept | ionic substitution probabilities — not defined in this paper's own words; probabilities adjusted to prioritize discovery Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[31] |
| concept | incomplete replacements — not defined in this paper's own words; partial substitutions enabled by SAPS Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[32] |
| concept | volume-based test-time augmentation — not defined in this paper's own words; GNoME filtration technique Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[33] |
| concept | uncertainty quantification through deep ensembles — not defined in this paper's own words; second GNoME filtration technique Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[34] |
| concept | polymorphs — not defined in this paper's own words; clustered structures ranked for DFT evaluation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[35] |
| concept | compositional models — not defined in this paper's own words; models predicting stability without structural information Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[36] |
| concept | reduced chemical formulas — not defined in this paper's own words; inputs to composition-only models Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[37] |
| concept | oxidation-state balancing — not defined in this paper's own words; composition-generation method described as overly strict Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[38] |
| concept | relaxed constraints — not defined in this paper's own words; looser constraints used for composition filtration Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[39] |
| concept | ab initio random structure searching (AIRSS) — not defined in this paper's own words; random structural evaluation of filtered compositions Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[40] |
| concept | relative stability (decomposition energy) — not defined in this paper's own words; thresholding quantity relative to competing phases Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[41] |
| concept | Vienna Ab initio Simulation Package (VASP) — not defined in this paper's own words; software used for DFT evaluation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[42] |
| concept | number of stable materials discovered — not defined in this paper's own words; discovery-output measure Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[43] |
| concept | total energy of a crystal — not defined in this paper's own words; prediction target of all GNoME models Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[44] |
| concept | one-hot embedding of the elements — not defined in this paper's own words; conversion from crystal inputs to graph features Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[45] |
| concept | message-passing formulation — not defined in this paper's own words; GNoME graph-network update formulation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[46] |
| concept | shallow multilayer perceptrons (MLPs) — not defined in this paper's own words; aggregate projections used during message passing Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[47] |
| concept | swish nonlinearities — not defined in this paper's own words; nonlinearity used by the MLPs Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[48] |
| concept | average adjacency of atoms — not defined in this paper's own words; dataset-wide normalizer for edge-to-node messages Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[49] |
| concept | mean absolute error (MAE) — not defined in this paper's own words; energy-prediction error benchmark Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[50] |
| concept | test loss performance — not defined in this paper's own words; quantity improving by a power law with more data Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[51] |
| concept | power law with further data — not defined in this paper's own words; reported scaling relationship for GNoME test loss Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[52] |
| concept | neural scaling laws — not defined in this paper's own words; deep-learning scaling regularity matched by GNoME Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[53] |
| concept | out-of-distribution tasks — not defined in this paper's own words; random-search inputs used to test emergent generalization Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[54] |
| concept | high-energy local minima — not defined in this paper's own words; random-search examples outside structural-pipeline distribution Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[55] |
| concept | universal energy predictor — not defined in this paper's own words; community model toward which scaled GNoME is a step Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[56] |
| concept | ionic relaxations — not defined in this paper's own words; source of diverse first-principles energies and forces Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[57] |
| concept | first-principles energies and forces — not defined in this paper's own words; reference quantities generated during relaxation Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[58] |
| concept | general-purpose machine-learning interatomic potential (MLIP) — not defined in this paper's own words; bulk-solids potential pretrained from the GNoME dataset Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[59] |
| concept | bulk solids — not defined in this paper's own words; material class targeted by the general-purpose potential Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[60] |
| concept | ab initio molecular dynamics (AIMD) — not defined in this paper's own words; typical per-material data source for prior potentials Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[61] |
| concept | general applicability and adoption — not defined in this paper's own words; properties limited by per-material training Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[62] |
| concept | large-scale pretraining of MLIPs — not defined in this paper's own words; pretraining that yields zero-shot models Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[63] |
| concept | unprecedented zero-shot accuracy — not defined in this paper's own words; performance enabled by large-scale pretraining Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[64] |
| concept | superionic conductors — not defined in this paper's own words; materials discoverable by the pretrained potentials Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[65] |
| concept | material-specific data — not defined in this paper's own words; training data not required for zero-shot discovery Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[66] |
| concept | increasing the pretraining dataset — not defined in this paper's own words; scaling intervention producing power-law accuracy gains Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[67] |
| concept | power-law improvements in accuracy — not defined in this paper's own words; accuracy trend under increasing pretraining scale Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[68] |
| concept | pretrained GNoME potential — not defined in this paper's own words; potential evaluated zero-shot on an AIMD distribution Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[69] |
| concept | unseen compositions — not defined in this paper's own words; zero-shot inputs absent from training Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[70] |
| concept | melted structures — not defined in this paper's own words; second unseen zero-shot structure class Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[71] |
| concept | structures including vacancies — not defined in this paper's own words; third unseen zero-shot structure class Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[72] |
| concept | transferability of MLIPs — not defined in this paper's own words; distribution-shift capability improved by pretraining Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[73] |
| concept | potential pretrained on GNoME data — not defined in this paper's own words; fine-tuned potential compared with training from scratch Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[74] |
| concept | potential trained from scratch — not defined in this paper's own words; comparison baseline for transferability Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[75] |
| concept | GNoME-pretrained — not defined in this paper's own words; pretrained potential in the figure comparison Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[76] |
| concept | randomly initialized potential — not defined in this paper's own words; non-pretrained comparison potential Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[77] |
| concept | distribution shift — not defined in this paper's own words; 400 K to 1,000 K robustness test Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[78] |
| concept | stable crystals — not defined in this paper's own words; discovery count increased by scaled GNNs Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[79] |
| concept | identically distributed data at train and test times — not defined in this paper's own words; standard learning assumption challenged by discovery Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[80] |
| concept | out-of-distribution effort — not defined in this paper's own words; character assigned to scientific discovery Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[81] |
| concept | emergent out-of-distribution capabilities — not defined in this paper's own words; capabilities demonstrated at GNoME scale Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[82] |
| concept | unseen chemical spaces — not defined in this paper's own words; novel compositional regions reached by GNoME Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[83] |
| concept | kinetic properties — not defined in this paper's own words; new downstream prediction task Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | concepts[84] |
| power-law improvement of test loss with data | Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.statistics[0] |
| precision of stable predictions, reported as a hit rate | Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | content_tags.statistics[1] |
| claim type | method — a second reading DIFFERING from the recorded value; recorded: existence Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | alternatives.claim_type |
| method | system-or-tool-building — a second reading DIFFERING from the recorded value; recorded: system-or-tool-building Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | alternatives.method |
| phenomenon | discovery of energetically stable inorganic crystal structures at a scale experiment cannot reach — a second reading DIFFERING from the recorded value; recorded: emergent out-of-distribution generalisation in a graph network trained at scale Quotation withheld — this paper is not ours to reproduce, and the page has already quoted as much of it as the rights layer allows. The tag, its field and its confidence stand; the sentence is at the publisher. | alternatives.phenomenon |
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.
Tag layer © the Meta-Tagging Project, CC BY 4.0. The paper text is © its authors/publisher
and is linked, not reproduced. Excerpts shown are short quotations for identification and verification.
· back to index
· concepts · authors
· עברית
© 2026 Shir Sivroni · ORCID 0009-0000-2597-2824