{"id":"https://openalex.org/W4412888798","doi":"https://doi.org/10.18653/v1/2025.findings-acl.159","title":"To Code or not to Code? Adaptive Tool Integration for Math Language Models via Expectation-Maximization","display_name":"To Code or not to Code? Adaptive Tool Integration for Math Language Models via Expectation-Maximization","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888798","doi":"https://doi.org/10.18653/v1/2025.findings-acl.159"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.159","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.159","pdf_url":"https://aclanthology.org/2025.findings-acl.159.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.159.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Haozhe Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haozhe Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100408767","display_name":"Long Li","orcid":"https://orcid.org/0000-0002-3746-463X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Long Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102815996","display_name":"Chao Qu","orcid":"https://orcid.org/0000-0002-7655-6318"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chao Qu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103257775","display_name":"Weidi Xu","orcid":"https://orcid.org/0000-0002-7279-9339"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Weidi Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113724064","display_name":"Fengming Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fengming Zhu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040527201","display_name":"Wei Chu","orcid":"https://orcid.org/0000-0002-7704-5544"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei Chu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5102011179","display_name":"Fangzhen Lin","orcid":"https://orcid.org/0000-0002-3141-8675"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fangzhen Lin","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":"3060","last_page":"3075"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9980999827384949,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9980999827384949,"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/T10028","display_name":"Topic Modeling","score":0.9886999726295471,"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/T11450","display_name":"Model-Driven Software Engineering Techniques","score":0.9828000068664551,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.677445650100708},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6599525809288025},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.5598823428153992},{"id":"https://openalex.org/keywords/maximization","display_name":"Maximization","score":0.545129656791687},{"id":"https://openalex.org/keywords/code-generation","display_name":"Code generation","score":0.42851293087005615},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.4032505750656128},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.23882073163986206},{"id":"https://openalex.org/keywords/mathematical-optimization","display_name":"Mathematical optimization","score":0.17717212438583374}],"concepts":[{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.677445650100708},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6599525809288025},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.5598823428153992},{"id":"https://openalex.org/C2776330181","wikidata":"https://www.wikidata.org/wiki/Q18358244","display_name":"Maximization","level":2,"score":0.545129656791687},{"id":"https://openalex.org/C133162039","wikidata":"https://www.wikidata.org/wiki/Q1061077","display_name":"Code generation","level":3,"score":0.42851293087005615},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.4032505750656128},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.23882073163986206},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.17717212438583374},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.0},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.0},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.159","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.159","pdf_url":"https://aclanthology.org/2025.findings-acl.159.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},{"id":"pmh:oai:repository.hkust.edu.hk:1783.1-168822","is_oa":false,"landing_page_url":"http://repository.hkust.edu.hk/ir/Record/1783.1-168822","pdf_url":null,"source":{"id":"https://openalex.org/S4306401796","display_name":"Rare & Special e-Zone (The Hong Kong University of Science and Technology)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I200769079","host_organization_name":"Hong Kong University of Science and Technology","host_organization_lineage":["https://openalex.org/I200769079"],"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":"Conference paper"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.159","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.159","pdf_url":"https://aclanthology.org/2025.findings-acl.159.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[{"display_name":"Quality Education","score":0.699999988079071,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888798.pdf","grobid_xml":"https://content.openalex.org/works/W4412888798.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2947806671","https://openalex.org/W2390878257","https://openalex.org/W2144260821","https://openalex.org/W2141163797","https://openalex.org/W2118922860","https://openalex.org/W4313429348","https://openalex.org/W2163672025","https://openalex.org/W2048831961","https://openalex.org/W1606349578","https://openalex.org/W4399567378"],"abstract_inverted_index":{"Recent":[0],"advances":[1],"in":[2,95],"mathematical":[3],"problemsolving":[4],"with":[5,128],"language":[6],"models":[7,66,168],"(LMs)":[8],"integrate":[9,55],"chain-of-thought":[10],"(CoT)":[11],"reasoning":[12,73,86],"and":[13,48,52,141,161,169],"code":[14,63,98],"execution":[15],"to":[16,43,54,67,101],"harness":[17],"their":[18,72],"complementary":[19],"strengths.However,":[20],"existing":[21],"hybrid":[22],"frameworks":[23],"exhibit":[24],"a":[25,118,134],"critical":[26],"limitation:":[27],"they":[28],"depend":[29],"on":[30,159,163,173],"externally":[31],"dictated":[32],"instructions":[33],"or":[34],"rigid":[35],"codeintegration":[36],"templates,":[37],"lacking":[38],"metacognitive":[39,138],"awareness":[40],"-the":[41],"capacity":[42],"dynamically":[44],"evaluate":[45],"intrinsic":[46],"capabilities":[47],"autonomously":[49],"determine":[50],"when":[51],"how":[53],"tools.This":[56],"rigidity":[57],"motivates":[58],"our":[59,145,153],"study":[60],"of":[61,104,109],"autonomous":[62,97],"integration,":[64],"enabling":[65],"adapt":[68],"tool-usage":[69],"strategies":[70],"as":[71],"abilities":[74],"evolve":[75],"during":[76],"training.While":[77],"reinforcement":[78],"learning":[79,96],"(RL)":[80],"shows":[81],"promise":[82],"for":[83],"boosting":[84],"LLM":[85],"at":[87],"scale":[88],"(e.g.,":[89],"DeepSeek-R1),":[90],"we":[91,116],"demonstrate":[92],"its":[93],"inefficiency":[94],"integration":[99],"due":[100],"inadequate":[102],"exploration":[103,126],"the":[105],"vast":[106],"combinatorial":[107],"space":[108],"CoT-code":[110],"interleaving":[111],"patterns.To":[112],"address":[113],"this":[114],"challenge,":[115],"propose":[117],"novel":[119],"Expectation-Maximization":[120],"(EM)":[121],"framework":[122],"that":[123],"synergizes":[124],"structured":[125],"(E-step)":[127],"off-policy":[129],"RL":[130],"optimization":[131],"(M-step),":[132],"creating":[133],"self-reinforcing":[135],"cycle":[136],"between":[137],"tool-use":[139],"decisions":[140],"evolving":[142],"capabilities.Experiments":[143],"reveal":[144],"method":[146],"achieves":[147],"superior":[148],"results":[149],"through":[150],"improved":[151],"exploration.Notably,":[152],"7B":[154],"model":[155],"improves":[156],"over":[157],"11%":[158],"MATH500":[160],"9.4%":[162],"AIME":[164],"without":[165],"o1-like":[166],"CoT.Code,":[167],"data":[170],"are":[171],"released":[172],"https://github.com/HaozheH3/AutoCode.":[174]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}