{"id":"https://openalex.org/W4412886726","doi":"https://doi.org/10.18653/v1/2025.acl-long.235","title":"Principled Understanding of Generalization for Generative Transformer Models in Arithmetic Reasoning Tasks","display_name":"Principled Understanding of Generalization for Generative Transformer Models in Arithmetic Reasoning Tasks","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412886726","doi":"https://doi.org/10.18653/v1/2025.acl-long.235"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.235","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.235","pdf_url":"https://aclanthology.org/2025.acl-long.235.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.235.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5049694379","display_name":"Xingcheng Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xingcheng Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039427876","display_name":"Zibo Zhao","orcid":"https://orcid.org/0000-0003-1089-2375"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zibo Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100758732","display_name":"Haipeng Zhang","orcid":"https://orcid.org/0000-0001-5741-2311"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haipeng Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5001935213","display_name":"Yanqing Yang","orcid":"https://orcid.org/0009-0006-9175-2137"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yanqing Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"4721","last_page":"4747"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11975","display_name":"Evolutionary Algorithms and Applications","score":0.8946999907493591,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11975","display_name":"Evolutionary Algorithms and Applications","score":0.8946999907493591,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11902","display_name":"Intelligent Tutoring Systems and Adaptive Learning","score":0.8529000282287598,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T14394","display_name":"Cognitive Science and Education Research","score":0.7997000217437744,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.6779878735542297},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.6740760207176208},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6202117204666138},{"id":"https://openalex.org/keywords/generative-grammar","display_name":"Generative grammar","score":0.5872502326965332},{"id":"https://openalex.org/keywords/arithmetic","display_name":"Arithmetic","score":0.465051531791687},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3652711510658264},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.3394366502761841},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.23105937242507935},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.06380012631416321}],"concepts":[{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.6779878735542297},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6740760207176208},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6202117204666138},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.5872502326965332},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.465051531791687},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3652711510658264},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3394366502761841},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.23105937242507935},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.06380012631416321},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.0},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.235","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.235","pdf_url":"https://aclanthology.org/2025.acl-long.235.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.235","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.235","pdf_url":"https://aclanthology.org/2025.acl-long.235.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412886726.pdf","grobid_xml":"https://content.openalex.org/works/W4412886726.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2380075625","https://openalex.org/W3162204513","https://openalex.org/W2371138613","https://openalex.org/W2048963458","https://openalex.org/W43109613","https://openalex.org/W2359952343","https://openalex.org/W2239445980","https://openalex.org/W2080152487","https://openalex.org/W3083152911","https://openalex.org/W3022347918"],"abstract_inverted_index":{"Transformer-based":[0],"models":[1,101],"excel":[2],"in":[3,11,34,39,60,80,93,138],"various":[4],"tasks":[5,17],"but":[6],"their":[7],"generalization":[8,38,56,110,137],"capabilities,":[9,25],"especially":[10],"arithmetic":[12,61],"reasoning,":[13],"remain":[14],"incompletely":[15],"understood.Arithmetic":[16],"provide":[18],"a":[19,49,130],"controlled":[20],"framework":[21,52],"to":[22,108,133],"explore":[23],"these":[24],"yet":[26],"performance":[27],"anomalies":[28],"persist,":[29],"such":[30],"as":[31],"inconsistent":[32],"effectiveness":[33],"multiplication":[35],"and":[36,72,119,126],"erratic":[37],"modular":[40,73,94],"addition":[41,81],"(e.g.,":[42],"modulo":[43],"100":[44],"vs.":[45],"101).This":[46],"paper":[47],"develops":[48],"unified":[50],"theoretical":[51],"for":[53,87,123],"understanding":[54,134],"the":[55,114],"behaviors":[57],"of":[58,69,116,135],"transformers":[59],"tasks,":[62],"focusing":[63],"on":[64],"length":[65,136],"generalization.Through":[66],"detailed":[67],"analysis":[68],"addition,":[70],"multiplication,":[71],"operations,":[74],"we":[75],"reveal":[76],"that":[77],"translation":[78],"invariance":[79],"aligns":[82],"with":[83],"relative":[84],"positional":[85],"encoding":[86],"robust":[88],"generalization,":[89],"while":[90],"base":[91],"mismatch":[92],"operations":[95],"disrupts":[96],"this":[97],"alignment.Experiments":[98],"across":[99],"GPT-family":[100],"validate":[102],"our":[103],"framework,":[104],"confirming":[105],"its":[106],"ability":[107],"predict":[109],"behaviors.Our":[111],"work":[112],"highlights":[113],"importance":[115],"task":[117],"structure":[118],"training":[120],"data":[121],"distribution":[122],"achieving":[124],"data-efficient":[125],"structure-aware":[127],"training,":[128],"providing":[129],"systematic":[131],"approach":[132],"transformers.":[139]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}