{"id":"https://openalex.org/W7202139487","doi":"https://doi.org/10.48550/arxiv.2608.06582","title":"CrystalGRPO: Target-Aligned and Coverage-Preserving Reinforcement Learning for Flow-Based Crystal Structure Prediction","display_name":"CrystalGRPO: Target-Aligned and Coverage-Preserving Reinforcement Learning for Flow-Based Crystal Structure Prediction","publication_year":2026,"publication_date":"2026-08-06","ids":{"openalex":"https://openalex.org/W7202139487","doi":"https://doi.org/10.48550/arxiv.2608.06582"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.06582","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.06582","pdf_url":"https://arxiv.org/pdf/2608.06582","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.06582","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137186740","display_name":"Kaixiang Su","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I102149020","display_name":"University of North Carolina at Charlotte","ror":"https://ror.org/04dawnj30","country_code":"US","type":"education","lineage":["https://openalex.org/I102149020"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Su, Kaixiang","raw_affiliation_strings":["Department of Computer Science , University of North Carolina at Charlotte"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science , University of North Carolina at Charlotte","institution_ids":["https://openalex.org/I102149020"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135263133","display_name":"Hongfei Xue","orcid":"https://orcid.org/0000-0001-9691-9668","observed_orcids":["https://orcid.org/0000-0001-9691-9668"]},"institutions":[{"id":"https://openalex.org/I102149020","display_name":"University of North Carolina at Charlotte","ror":"https://ror.org/04dawnj30","country_code":"US","type":"education","lineage":["https://openalex.org/I102149020"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Xue, Hongfei","raw_affiliation_strings":["Department of Computer Science , University of North Carolina at Charlotte"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science , University of North Carolina at Charlotte","institution_ids":["https://openalex.org/I102149020"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5146394996","display_name":"Qiang Zhu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I102149020","display_name":"University of North Carolina at Charlotte","ror":"https://ror.org/04dawnj30","country_code":"US","type":"education","lineage":["https://openalex.org/I102149020"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhu, Qiang","raw_affiliation_strings":["Department of Mechanical Engineering and Engineering Science , University of North Carolina at Charlotte","North Carolina Battery Complexity , Autonomous Vehicle and Electrification (BATT CAVE) Research Center Charlotte , NC 28223 , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Mechanical Engineering and Engineering Science , University of North Carolina at Charlotte","institution_ids":["https://openalex.org/I102149020"]},{"raw_affiliation_string":"North Carolina Battery Complexity , Autonomous Vehicle and Electrification (BATT CAVE) Research Center Charlotte , NC 28223 , USA","institution_ids":["https://openalex.org/I102149020"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I102149020"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":2,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11948","display_name":"Machine Learning in Materials Science","score":0.9990000128746033,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11948","display_name":"Machine Learning in Materials Science","score":0.9990000128746033,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10211","display_name":"Computational Drug Discovery Methods","score":0.00039999998989515007,"subfield":{"id":"https://openalex.org/subfields/3002","display_name":"Drug Discovery"},"field":{"id":"https://openalex.org/fields/30","display_name":"Pharmacology, Toxicology and Pharmaceutics"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T10002","display_name":"Advanced Chemical Physics Studies","score":0.00019999999494757503,"subfield":{"id":"https://openalex.org/subfields/3107","display_name":"Atomic and Molecular Physics, and Optics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/crystal-structure-prediction","display_name":"crystal structure prediction","score":0.9919999837875366},{"id":"https://openalex.org/keywords/flow-based-generative-models","display_name":"flow-based generative models","score":0.9470000267028809},{"id":"https://openalex.org/keywords/reinforcement-learning-post-training","display_name":"reinforcement learning post-training","score":0.8949999809265137},{"id":"https://openalex.org/keywords/target-recovery","display_name":"target recovery","score":0.5960000157356262},{"id":"https://openalex.org/keywords/coverage-preservation","display_name":"coverage preservation","score":0.5690000057220459}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5805000066757202},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5723000168800354},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.5458999872207642},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4717000126838684},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.45809999108314514},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.4422000050544739},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.43220001459121704},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4072999954223633},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.3781999945640564},{"id":"https://openalex.org/C45804977","wikidata":"https://www.wikidata.org/wiki/Q7239673","display_name":"Predictive modelling","level":2,"score":0.3508000075817108},{"id":"https://openalex.org/C18555067","wikidata":"https://www.wikidata.org/wiki/Q8375051","display_name":"Joint (building)","level":2,"score":0.34940001368522644},{"id":"https://openalex.org/C2781311116","wikidata":"https://www.wikidata.org/wiki/Q83306","display_name":"Group (periodic table)","level":2,"score":0.34779998660087585},{"id":"https://openalex.org/C84947059","wikidata":"https://www.wikidata.org/wiki/Q5191417","display_name":"Crystal structure prediction","level":3,"score":0.33239999413490295},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.320499986410141},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.31859999895095825},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3118000030517578},{"id":"https://openalex.org/C2777115002","wikidata":"https://www.wikidata.org/wiki/Q7168246","display_name":"Performance prediction","level":2,"score":0.29109999537467957},{"id":"https://openalex.org/C139945424","wikidata":"https://www.wikidata.org/wiki/Q1940696","display_name":"Mean squared error","level":2,"score":0.265500009059906},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2574999928474426}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.06582","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.06582","pdf_url":"https://arxiv.org/pdf/2608.06582","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.06582","doi":"https://doi.org/10.48550/arxiv.2608.06582","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.06582","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.06582","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.06582","pdf_url":"https://arxiv.org/pdf/2608.06582","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7202139487.pdf","grobid_xml":"https://content.openalex.org/works/W7202139487.grobid-xml"},"referenced_works_count":7,"referenced_works":["https://openalex.org/W2736601468","https://openalex.org/W4378464471","https://openalex.org/W4391631327","https://openalex.org/W4391631952","https://openalex.org/W4392182166","https://openalex.org/W4407185557","https://openalex.org/W7112195364"],"related_works":["https://openalex.org/W7215122725","https://openalex.org/W7212166952","https://openalex.org/W4387045209","https://openalex.org/W7117152685","https://openalex.org/W7131386686","https://openalex.org/W4412609221","https://openalex.org/W7197021136","https://openalex.org/W7214108720","https://openalex.org/W7220655348","https://openalex.org/W7130762369"],"abstract_inverted_index":{"Flow-based":[0],"generative":[1],"models":[2],"can":[3,53],"efficiently":[4],"produce":[5],"candidate":[6,56],"structures":[7],"for":[8,59],"crystal":[9],"structure":[10],"prediction":[11],"(CSP),":[12],"but":[13],"their":[14],"pretrained":[15],"objectives":[16],"do":[17],"not":[18,45],"directly":[19],"optimize":[20],"downstream":[21],"target":[22,114],"recovery.":[23,61,115],"Reinforcement-learning":[24],"post-training":[25,67],"offers":[26],"a":[27,65,85,107,148],"flexible":[28],"solution,":[29],"yet":[30],"existing":[31,71],"approaches":[32],"rely":[33],"primarily":[34],"on":[35],"energy":[36,43,83],"rewards":[37],"and":[38,89,99,118,122,129],"coordinate-only":[39,134],"stochastic":[40],"policies.":[41],"Predicted":[42],"does":[44],"identify":[46],"the":[47,55,76],"reference":[48,104],"polymorph,":[49],"while":[50],"reward-driven":[51],"concentration":[52],"reduce":[54,127],"coverage":[57],"required":[58],"Top-N":[60],"We":[62],"introduce":[63],"CrystalGRPO,":[64],"CSP-aligned":[66],"framework":[68],"that":[69],"extends":[70],"ODE-to-SDE":[72],"policy":[73],"constructions":[74],"to":[75,111,133],"joint":[77],"coordinate--lattice":[78],"state.":[79],"CrystalGRPO":[80],"combines":[81,102],"MACE-predicted":[82],"with":[84,106,120],"StructureMatcher-based":[86],"recovery":[87],"score":[88],"provides":[90],"two":[91],"operating":[92],"modes:":[93],"CrystalGRPO-Q,":[94],"which":[95,101],"prioritizes":[96],"single-draw":[97],"recovery,":[98],"CrystalGRPO-C,":[100],"full-trajectory":[103],"regularization":[105],"coverage-aware":[108],"group":[109],"advantage":[110],"preserve":[112],"finite-budget":[113],"Across":[116],"MP-20":[117],"MPTS-52":[119],"PXRDGen":[121],"OMatG":[123],"backbones,":[124],"both":[125],"variants":[126],"one-":[128],"twenty-sample":[130],"RMSE":[131],"relative":[132],"reinforcement":[135],"in":[136],"all":[137,152],"four":[138],"backbone--dataset":[139],"settings.":[140,153],"CrystalGRPO-Q":[141],"consistently":[142],"improves":[143],"Top-1,":[144],"whereas":[145],"CrystalGRPO-C":[146],"achieves":[147],"higher":[149],"Top-20":[150],"across":[151]},"counts_by_year":[{"year":2026,"cited_by_count":2}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-08-11T00:00:00"}