{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,12]],"date-time":"2026-10-12T10:47:25Z","timestamp":1791802045604,"version":"4.3.4"},"reference-count":55,"publisher":"AIP Publishing","issue":"14","license":[{"start":{"date-parts":[[2026,10,12]],"date-time":"2026-10-12T00:00:00Z","timestamp":1791763200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100001711","name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung","doi-asserted-by":"publisher","award":["200020_214879"],"award-info":[{"award-number":["200020_214879"]}],"id":[{"id":"10.13039\/501100001711","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["pubs.aip.org"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2026,10,14]]},"abstract":"<jats:p>Mapping an atomic structure to a compact set of geometric descriptors is an essential step in any machine-learning application to atomic-scale modeling. A powerful and widely used approach can be understood as a discretization of the histogram of pair distances, triangles, etc., which results in a hierarchy of symmetry-invariant atom-centered descriptors. Unfortunately, the lower rungs on this hierarchy (two, three, four-body clusters) were found to be incomplete, with symmetry-unrelated pairs of structures having exactly the same descriptors. However, all the \u201cdegeneracies\u201d reported so far are resolved by considering larger clusters of neighbors to build the descriptors. We report examples of 3D structures that are indistinguishable even if one considers clusters of up to seven neighbors, and to arbitrary order when considering a practical level of discretization of the descriptors, discovered with the assistance of large language models. The key ingredients in their construction can be traced to results that have been known for decades in different communities: the model was able to find the references and recognize their significance for the problem at hand. We believe that this experiment exposes an extremely fruitful usage pattern for Artificial Intelligence in science: translating results between different communities and application domains, accelerating the process by which serendipitous discoveries in a field become breakthroughs in another.<\/jats:p>","DOI":"10.1063\/5.0354408","type":"journal-article","created":{"date-parts":[[2026,10,12]],"date-time":"2026-10-12T10:10:52Z","timestamp":1791799852000},"update-policy":"https:\/\/doi.org\/10.1063\/aip-crossmark-policy-page","source":"Crossref","is-referenced-by-count":0,"title":["Using large language models to probe the limits of atom-centered structural descriptors"],"prefix":"10.1063","volume":"165","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-7009-9240","authenticated-orcid":false,"given":"Michelangelo","family":"Domina","sequence":"first","affiliation":[{"name":"Laboratory of Computational Science and Modeling, Institut des Mat\u00e9riaux, \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne , 1015 Lausanne,","place":["Switzerland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2571-2832","authenticated-orcid":false,"given":"Michele","family":"Ceriotti","sequence":"additional","affiliation":[{"name":"Laboratory of Computational Science and Modeling, Institut des Mat\u00e9riaux, \u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne , 1015 Lausanne,","place":["Switzerland"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"317","published-online":{"date-parts":[[2026,10,12]]},"reference":[{"key":"2026101206104951100_c1","doi-asserted-by":"publisher","first-page":"146401","DOI":"10.1103\/physrevlett.98.146401","article-title":"Generalized neural-network representation of high-dimensional potential-energy surfaces","volume":"98","year":"2007","journal-title":"Phys. Rev. Lett."},{"key":"2026101206104951100_c2","doi-asserted-by":"publisher","first-page":"058301","DOI":"10.1103\/physrevlett.108.058301","article-title":"Fast and accurate modeling of molecular atomization energies with machine learning","volume":"108","year":"2012","journal-title":"Phys. Rev. Lett."},{"key":"2026101206104951100_c3","doi-asserted-by":"publisher","first-page":"136403","DOI":"10.1103\/physrevlett.104.136403","article-title":"Gaussian approximation potentials: The accuracy of quantum mechanics, without the electrons","volume":"104","year":"2010","journal-title":"Phys. Rev. Lett."},{"key":"2026101206104951100_c4","doi-asserted-by":"publisher","first-page":"9759","DOI":"10.1021\/acs.chemrev.1c00021","article-title":"Physics-inspired structural representations for molecules and materials","volume":"121","year":"2021","journal-title":"Chem. Rev."},{"key":"2026101206104951100_c5","doi-asserted-by":"publisher","first-page":"10037","DOI":"10.1021\/acs.chemrev.0c00868","article-title":"Four generations of high-dimensional neural network potentials","volume":"121","year":"2021","journal-title":"Chem. Rev."},{"key":"2026101206104951100_c6","doi-asserted-by":"publisher","first-page":"10142","DOI":"10.1021\/acs.chemrev.0c01111","article-title":"Machine learning force fields","volume":"121","year":"2021","journal-title":"Chem. Rev."},{"key":"2026101206104951100_c7","doi-asserted-by":"publisher","first-page":"57","DOI":"10.1021\/acscentsci.8b00551","article-title":"Transferable machine-learning model of the electron density","volume":"5","year":"2019","journal-title":"ACS Cent. Sci."},{"key":"2026101206104951100_c8","doi-asserted-by":"publisher","first-page":"6772","DOI":"10.1038\/s41467-024-51006-4","article-title":"Deep learning tight-binding approach for large-scale electronic simulations at finite temperatures with ab\u00a0initio accuracy","volume":"15","year":"2024","journal-title":"Nat. Commun."},{"key":"2026101206104951100_c9","doi-asserted-by":"publisher","first-page":"154110","DOI":"10.1063\/1.5090481","article-title":"Atom-density representations for machine learning","volume":"150","year":"2019","journal-title":"J.\u00a0Chem. Phys."},{"key":"2026101206104951100_c10","doi-asserted-by":"publisher","first-page":"166001","DOI":"10.1103\/physrevlett.125.166001","article-title":"Incompleteness of atomic structure representations","volume":"125","year":"2020","journal-title":"Phys. Rev. Lett."},{"key":"2026101206104951100_c11","doi-asserted-by":"publisher","first-page":"126","DOI":"10.12688\/openreseurope.14156.1","article-title":"Local invertibility and sensitivity of atomic structure-feature mappings","volume":"1","year":"2021","journal-title":"Open Res. Eur."},{"key":"2026101206104951100_c12","doi-asserted-by":"publisher","first-page":"045020","DOI":"10.1088\/2632-2153\/aca1f8","article-title":"Incompleteness of graph neural networks for points clouds in three dimensions","volume":"3","year":"2022","journal-title":"Mach. Learn.: Sci. Technol."},{"key":"2026101206104951100_c13","doi-asserted-by":"publisher","first-page":"034302","DOI":"10.1063\/5.0070488","article-title":"Manifolds of quasi-constant SOAP and ACSF fingerprints and the resulting failure to machine learn four-body interactions","volume":"156","year":"2022","journal-title":"J.\u00a0Chem. Phys."},{"key":"2026101206104951100_c14","doi-asserted-by":"publisher","first-page":"177101","DOI":"10.1063\/5.0088404","article-title":"Comment on \u2018Manifolds of quasi-constant SOAP and ACSF fingerprints and the resulting failure to machine learn four-body interactions\u2019 [J. Chem. Phys. 156, 034302 (2022)]","volume":"157","year":"2022","journal-title":"J.\u00a0Chem. Phys."},{"key":"2026101206104951100_c15","doi-asserted-by":"publisher","first-page":"709","DOI":"10.1016\/s0196-8858(03)00101-5","article-title":"On reconstructing n-point configurations from the distribution of distances or areas","volume":"32","year":"2004","journal-title":"Adv. Appl. Math."},{"key":"2026101206104951100_c16","doi-asserted-by":"publisher","first-page":"014104","DOI":"10.1103\/physrevb.99.014104","article-title":"Atomic cluster expansion for accurate and transferable interatomic potentials","volume":"99","year":"2019","journal-title":"Phys. Rev. B"},{"key":"2026101206104951100_c17","doi-asserted-by":"publisher","first-page":"121101","DOI":"10.1063\/5.0021116","article-title":"Recursive evaluation and iterative contraction of N-body equivariant features","volume":"153","year":"2020","journal-title":"J.\u00a0Chem. Phys."},{"key":"2026101206104951100_c18","doi-asserted-by":"publisher","first-page":"110946","DOI":"10.1016\/j.jcp.2022.110946","article-title":"Atomic cluster expansion: Completeness, efficiency and stability","volume":"454","year":"2022","journal-title":"J.\u00a0Comput. Phys."},{"key":"2026101206104951100_c19","doi-asserted-by":"publisher","first-page":"016110","DOI":"10.1063\/5.0160740","article-title":"Completeness of atomic structure representations","volume":"2","year":"2024","journal-title":"APL Mach. Learn."},{"key":"2026101206104951100_c20","doi-asserted-by":"publisher","first-page":"12513","DOI":"10.1021\/acs.jpclett.4c02376","article-title":"Complete and efficient covariants for three-dimensional point configurations with application to learning molecular quantum properties","volume":"15","year":"2024","journal-title":"J.\u00a0Phys. Chem. Lett."},{"key":"2026101206104951100_c21","doi-asserted-by":"publisher","first-page":"11423","DOI":"10.52202\/068431-0830","article-title":"MACE: Higher order equivariant message passing neural networks for fast and accurate force fields","volume":"35","year":"2022","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"2026101206104951100_c22","doi-asserted-by":"publisher","first-page":"2453","DOI":"10.1038\/s41467-022-29939-5","article-title":"E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials","volume":"13","year":"2022","journal-title":"Nat. Commun."},{"key":"2026101206104951100_c23","doi-asserted-by":"publisher","first-page":"79469","DOI":"10.52202\/075280-3478","article-title":"Smooth, exact rotational symmetrization for deep learning on point clouds","volume":"36","year":"2023","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"2026101206104951100_c24","doi-asserted-by":"publisher","first-page":"139030","DOI":"10.52202\/079017-4412","article-title":"The importance of being scalable: Improving the speed and accuracy of neural network interatomic potentials across chemical domains","volume":"37","year":"2024","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"2026101206104951100_c25","unstructured":"M.\n            Balko\n          , J.Greb\u00edk, P.Hub\u00e1\u010dek, M.Kouteck\u00fd, M.Kripner, V.Rozho\u0148, R.\u0160\u00e1mal, and A.Z\u00e1me\u010dn\u00edk, \u201cBolzano: Case studies in LLM-assisted mathematical research,\u201d arXiv:2604.16989 [cs.CL] (2026)."},{"key":"2026101206104951100_c26","unstructured":"T.\n            Feng\n          , T.Trinh, G.Bingham, J.Kang, S.Zhang, S.-h.Kim, K.Barreto, C.Schildkraut, J.Jung, J.Seo, C.Pagano, Y.Chervonyi, D.Hwang, K.Hou, S.Gukov, C.-C.Tsai, H.Choi, Y.Jin, W.-Y.Li, H.-A.Wu, R.-A.Shiu, Y.-S.Shih, Q. V.Le, and T.Luong, \u201cSemi-autonomous mathematics discovery with Gemini: A case study on the Erd\u0151s problems,\u201d arXiv:2601.22401 [cs.AI] (2026)."},{"key":"2026101206104951100_c27","unstructured":"W.\n            Sawin\n          , \u201cAn explicit lower bound for the unit distance problem,\u201d arXiv:2605.20579 [math.CO] (2026)."},{"key":"2026101206104951100_c28","doi-asserted-by":"publisher","first-page":"184115","DOI":"10.1103\/physrevb.87.184115","article-title":"On representing chemical environments","volume":"87","year":"2013","journal-title":"Phys. Rev. B"},{"key":"2026101206104951100_c29","doi-asserted-by":"publisher","first-page":"29661","DOI":"10.1039\/c8cp05921g","article-title":"Feature optimization for atomistic machine learning yields a data-driven construction of the periodic table of the elements","volume":"20","year":"2018","journal-title":"Phys. Chem. Chem. Phys."},{"key":"2026101206104951100_c30","doi-asserted-by":"publisher","first-page":"316","DOI":"10.1016\/j.jcp.2014.12.018","article-title":"Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials","volume":"285","year":"2015","journal-title":"J.\u00a0Comput. Phys."},{"key":"2026101206104951100_c31","doi-asserted-by":"publisher","first-page":"085106","DOI":"10.1103\/g49f-b1x5","article-title":"General formalism for machine-learning models based on multipolar spherical harmonics","volume":"112","year":"2025","journal-title":"Phys. Rev. B"},{"key":"2026101206104951100_c32","doi-asserted-by":"publisher","first-page":"341","DOI":"10.1007\/s10851-012-0330-6","article-title":"The bispectrum as a source of phase-sensitive invariants for Fourier descriptors: A group-theoretic approach","volume":"44","year":"2012","journal-title":"J.\u00a0Math. Imaging Vis."},{"key":"2026101206104951100_c33","doi-asserted-by":"publisher","first-page":"939","DOI":"10.1038\/143939b0","article-title":"Homometric structures","volume":"143","year":"1939","journal-title":"Nature"},{"key":"2026101206104951100_c34","doi-asserted-by":"publisher","first-page":"343","DOI":"10.1137\/0603035","article-title":"The structure of homometric sets","volume":"3","year":"1982","journal-title":"SIAM J. Algebraic Discrete Methods"},{"key":"2026101206104951100_c35","doi-asserted-by":"publisher","first-page":"074106","DOI":"10.1063\/1.3553717","article-title":"Atom-centered symmetry functions for constructing high-dimensional neural network potentials","volume":"134","year":"2011","journal-title":"J.\u00a0Chem. Phys."},{"key":"2026101206104951100_c36","article-title":"Validation scripts and input structures for \u2018Using large language models to probe the limits of atom-centered structural descriptors","year":"2026"},{"key":"2026101206104951100_c37","doi-asserted-by":"publisher","first-page":"79","DOI":"10.46793\/match.91-1.079k","article-title":"Polynomial-time algorithms for continuous metrics on atomic clouds of unordered points","volume":"91","year":"2024","journal-title":"MATCH Commun. Math. Comput. Chem."},{"key":"2026101206104951100_c38","doi-asserted-by":"publisher","first-page":"379","DOI":"10.46793\/match.97-2.11426","article-title":"Complete invariants of atomic clouds under rigid motion with Lipschitz continuous metrics in a polynomial time","volume":"97","year":"2027","journal-title":"MATCH Commun. Math. Comput. Chem."},{"key":"2026101206104951100_c39","doi-asserted-by":"publisher","first-page":"24625","DOI":"10.52202\/068431-1788","article-title":"Resolving the data ambiguity for periodic crystals","volume":"35","year":"2022","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"2026101206104951100_c40","doi-asserted-by":"publisher","first-page":"12482","DOI":"10.1609\/aaai.v38i11.29141","article-title":"Complete neural networks for complete Euclidean graphs","volume":"38","year":"2024","journal-title":"Proc. AAAI Conf. Artif. Intell."},{"key":"2026101206104951100_c41","doi-asserted-by":"publisher","first-page":"169","DOI":"10.1006\/jcta.1998.2870","article-title":"Reconstructing subsets of Zn","volume":"83","year":"1998","journal-title":"J.\u00a0Combin. Theory Ser. A"},{"key":"2026101206104951100_c42","doi-asserted-by":"publisher","first-page":"310","DOI":"10.1107\/s0108767394009827","article-title":"The use of higher-order invariants in the determination of generalized Patterson cyclotomic sets","volume":"51","year":"1995","journal-title":"Acta Crystallogr., Sect. A: Found. Crystallogr."},{"key":"2026101206104951100_c43","first-page":"63","article-title":"Uniqueness of pitch class spaces, minimal bases and Z partners","volume-title":"Diderot Forum on Mathematics and Music: Computational and Mathematical Methods in Music","author":"Feichtinger","year":"1999"},{"key":"2026101206104951100_c44","doi-asserted-by":"publisher","first-page":"99","DOI":"10.1080\/17459737.2011.608820","article-title":"Discrete phase retrieval in musical structures","volume":"5","year":"2011","journal-title":"J.\u00a0Math. Music"},{"key":"2026101206104951100_c45","doi-asserted-by":"publisher","first-page":"867","DOI":"10.1017\/s0963548303005807","article-title":"The reconstructibility of finite abelian groups","volume":"13","year":"2004","journal-title":"Comb. Probab. Comput."},{"key":"2026101206104951100_c46","doi-asserted-by":"publisher","first-page":"388","DOI":"10.1364\/josaa.9.000388","article-title":"Uniqueness properties of higher-order autocorrelation functions","volume":"9","year":"1992","journal-title":"J.\u00a0Opt. Soc. Am. A"},{"key":"2026101206104951100_c47","doi-asserted-by":"publisher","first-page":"83","DOI":"10.1080\/17459737.2011.608819","article-title":"Z-relation and homometry in musical distributions","volume":"5","year":"2011","journal-title":"J.\u00a0Math. Music"},{"key":"2026101206104951100_c48","doi-asserted-by":"publisher","first-page":"1641","DOI":"10.1038\/s41467-023-36741-4","article-title":"Surprising combinations of research contents and contexts are related to impact and emerge with scientific outsiders from distant disciplines","volume":"14","year":"2023","journal-title":"Nat. Commun."},{"key":"2026101206104951100_c49","doi-asserted-by":"publisher","first-page":"914","DOI":"10.1038\/s41586-026-10265-5","article-title":"Towards end-to-end automation of AI research","volume":"651","year":"2026","journal-title":"Nature"},{"issue":"122","key":"2026101206104951100_c50","doi-asserted-by":"publisher","first-page":"10380","DOI":"10.21105\/joss.10380","article-title":"Chemiscope 1.0: Interactive exploration of atomistic data from analysis to dissemination","volume":"11","year":"2026","journal-title":"J. Open Source Software"},{"key":"2026101206104951100_c51","doi-asserted-by":"publisher","first-page":"1643","DOI":"10.1137\/25m1733574","article-title":"Complete and continuous invariants of 1-periodic sequences in polynomial time","volume":"7","year":"2025","journal-title":"SIAM J. Math. Data Sci."},{"key":"2026101206104951100_c52","unstructured":"J.\n            Simon\n          , \u201cReconstructing colourings of finite groups,\u201d Ph.D. thesis, RWTH Aachen University, Aachen, Germany, 2017."},{"key":"2026101206104951100_c53","doi-asserted-by":"publisher","first-page":"503","DOI":"10.1017\/s0963548306007875","article-title":"Reconstructing odd necklaces","volume":"16","year":"2007","journal-title":"Comb. Probab. Comput."},{"key":"2026101206104951100_c54","doi-asserted-by":"publisher","first-page":"621","DOI":"10.1007\/s10955-009-9742-0","article-title":"How model sets can be determined by their two-point and three-point correlations","volume":"135","year":"2009","journal-title":"J.\u00a0Stat. Phys."},{"key":"2026101206104951100_c55","doi-asserted-by":"publisher","first-page":"487","DOI":"10.4153\/cmb-2011-106-5","article-title":"Weighted model sets and their higher point-correlations","volume":"55","year":"2012","journal-title":"Can. Math. Bull."}],"container-title":["The Journal of Chemical Physics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pubs.aip.org\/aip\/jcp\/article-pdf\/doi\/10.1063\/5.0354408\/21168037\/144108_1_5.0354408.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/pubs.aip.org\/aip\/jcp\/article-pdf\/doi\/10.1063\/5.0354408\/21168037\/144108_1_5.0354408.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,12]],"date-time":"2026-10-12T10:10:54Z","timestamp":1791799854000},"score":1,"resource":{"primary":{"URL":"https:\/\/pubs.aip.org\/jcp\/article\/165\/14\/144108\/3406359\/Using-large-language-models-to-probe-the-limits-of"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,12]]},"references-count":55,"journal-issue":{"issue":"14","published-print":{"date-parts":[[2026,10,14]]}},"URL":"https:\/\/doi.org\/10.1063\/5.0354408","relation":{},"ISSN":["0021-9606","1089-7690"],"issn-type":[{"value":"0021-9606","type":"print"},{"value":"1089-7690","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026,10,14]]},"published":{"date-parts":[[2026,10,12]]},"article-number":"144108"}}