{"id":"https://openalex.org/W7166804896","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1420","title":"From Implicit to Explicit: Token-Efficient Logical Supervision for Mathematical Reasoning in LLMs","display_name":"From Implicit to Explicit: Token-Efficient Logical Supervision for Mathematical Reasoning in LLMs","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166804896","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1420"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.1420","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1420","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1420","pdf_url":"https://aclanthology.org/2026.findings-acl.1420.pdf","source":{"id":"https://openalex.org/S4363605144","display_name":"Findings of the Association for Computational Linguistics: ACL (Findings of ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 2026","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2026.findings-acl.1420.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5122314593","display_name":"Shaojie Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405261251","display_name":"The Hong Kong University of Science and Technology (Guangzhou)","ror":"https://ror.org/050h0vm43","country_code":"CN","type":"education","lineage":["https://openalex.org/I4405261251"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shaojie Wang","raw_affiliation_strings":["Hong Kong University of Science and Technology (Guangzhou )"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hong Kong University of Science and Technology (Guangzhou )","institution_ids":["https://openalex.org/I4405261251"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5139744272","display_name":"Liang Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405261251","display_name":"The Hong Kong University of Science and Technology (Guangzhou)","ror":"https://ror.org/050h0vm43","country_code":"CN","type":"education","lineage":["https://openalex.org/I4405261251"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liang Zhang","raw_affiliation_strings":["Hong Kong University of Science and Technology (Guangzhou )"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hong Kong University of Science and Technology (Guangzhou )","institution_ids":["https://openalex.org/I4405261251"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I4405261251"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.43890353,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"28486","last_page":"28500"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.7095000147819519,"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/T10028","display_name":"Topic Modeling","score":0.7095000147819519,"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/T12522","display_name":"Mathematics Education and Pedagogy","score":0.0471000000834465,"subfield":{"id":"https://openalex.org/subfields/2604","display_name":"Applied Mathematics"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.03519999980926514,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.7680000066757202},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.640999972820282},{"id":"https://openalex.org/keywords/mathematical-problem-solving","display_name":"mathematical problem solving","score":0.5690000057220459},{"id":"https://openalex.org/keywords/out-of-distribution-generalization","display_name":"out-of-distribution generalization","score":0.7129999995231628}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4627000093460083},{"id":"https://openalex.org/C195344581","wikidata":"https://www.wikidata.org/wiki/Q2555318","display_name":"Automated reasoning","level":2,"score":0.42160001397132874},{"id":"https://openalex.org/C97364631","wikidata":"https://www.wikidata.org/wiki/Q484284","display_name":"Deductive reasoning","level":2,"score":0.4203000068664551},{"id":"https://openalex.org/C43971567","wikidata":"https://www.wikidata.org/wiki/Q3142865","display_name":"Logical reasoning","level":2,"score":0.4124999940395355},{"id":"https://openalex.org/C539667460","wikidata":"https://www.wikidata.org/wiki/Q2414942","display_name":"Management science","level":1,"score":0.3921000063419342},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3450999855995178},{"id":"https://openalex.org/C47884741","wikidata":"https://www.wikidata.org/wiki/Q1166618","display_name":"Mathematical logic","level":2,"score":0.3368000090122223},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.3314000070095062},{"id":"https://openalex.org/C111472728","wikidata":"https://www.wikidata.org/wiki/Q9471","display_name":"Epistemology","level":1,"score":0.321399986743927},{"id":"https://openalex.org/C20162079","wikidata":"https://www.wikidata.org/wiki/Q1151406","display_name":"Case-based reasoning","level":2,"score":0.32019999623298645},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.31049999594688416},{"id":"https://openalex.org/C2776325391","wikidata":"https://www.wikidata.org/wiki/Q6917865","display_name":"Motivated reasoning","level":3,"score":0.29100000858306885},{"id":"https://openalex.org/C134752490","wikidata":"https://www.wikidata.org/wiki/Q374182","display_name":"Logical consequence","level":2,"score":0.26969999074935913},{"id":"https://openalex.org/C21847791","wikidata":"https://www.wikidata.org/wiki/Q191081","display_name":"Logical conjunction","level":2,"score":0.26499998569488525},{"id":"https://openalex.org/C2992562121","wikidata":"https://www.wikidata.org/wiki/Q3817808","display_name":"Scientific reasoning","level":2,"score":0.2635999917984009},{"id":"https://openalex.org/C159032336","wikidata":"https://www.wikidata.org/wiki/Q2488768","display_name":"Non-monotonic logic","level":2,"score":0.26269999146461487},{"id":"https://openalex.org/C20693621","wikidata":"https://www.wikidata.org/wiki/Q6667502","display_name":"Logical framework","level":2,"score":0.2574000060558319},{"id":"https://openalex.org/C89288958","wikidata":"https://www.wikidata.org/wiki/Q7301504","display_name":"Reasoning system","level":2,"score":0.2554999887943268},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.25209999084472656}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.1420","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1420","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1420","pdf_url":"https://aclanthology.org/2026.findings-acl.1420.pdf","source":{"id":"https://openalex.org/S4363605144","display_name":"Findings of the Association for Computational Linguistics: ACL (Findings of ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.1420","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1420","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1420","pdf_url":"https://aclanthology.org/2026.findings-acl.1420.pdf","source":{"id":"https://openalex.org/S4363605144","display_name":"Findings of the Association for Computational Linguistics: ACL (Findings of ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166804896.pdf","grobid_xml":"https://content.openalex.org/works/W7166804896.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4416036043","https://openalex.org/W4416036511","https://openalex.org/W7116869808","https://openalex.org/W4412944855","https://openalex.org/W4386554674","https://openalex.org/W4416615452","https://openalex.org/W4403456064","https://openalex.org/W4402953080","https://openalex.org/W4405032217","https://openalex.org/W7138862905"],"abstract_inverted_index":{"Recent":[0],"studies":[1],"reveal":[2,42],"that":[3,43,87,148],"large":[4],"language":[5],"models":[6,117,144],"(LLMs)":[7],"exhibit":[8],"limited":[9],"logical":[10,29,39,81,109,127],"reasoning":[11],"abilities":[12],"in":[13],"mathematical":[14],"problem-solving,":[15],"instead":[16],"often":[17],"relying":[18],"on":[19,28,118],"pattern-matching":[20],"and":[21,41,96,145,156,164,172],"memorization.We":[22],"systematically":[23],"analyze":[24],"this":[25,47,68,119],"limitation,":[26],"focusing":[27],"relationship":[30,82,128],"understanding,":[31,129],"which":[32,92,97,132],"is":[33,86],"a":[34,76],"core":[35],"capability":[36,48],"underlying":[37],"genuine":[38],"reasoning,":[40],"errors":[44],"related":[45],"to":[46,62,94,99,107],"account":[49],"for":[50,126],"over":[51,178],"90%":[52],"of":[53,162],"incorrect":[54],"predictions,":[55],"with":[56,159],"Chain-of-Thought":[57],"Supervised":[58],"Fine-Tuning":[59],"(CoT-SFT)":[60],"failing":[61],"substantially":[63],"reduce":[64],"these":[65],"errors.To":[66],"address":[67],"bottleneck,":[69],"we":[70],"propose":[71],"First-Step":[72],"Logical":[73],"Reasoning":[74],"(FSLR),":[75],"lightweight":[77],"training":[78,116,171,174],"framework":[79],"targeting":[80],"understanding.Our":[83],"key":[84],"insight":[85],"the":[88,105,113],"first":[89],"planning":[90],"step-identifying":[91],"variables":[93],"use":[95],"operation":[98],"apply":[100],"without":[101],"performing":[102],"any":[103],"calculation-encourages":[104],"model":[106],"derive":[108],"relationships":[110,136],"directly":[111],"from":[112],"problem":[114],"statement.By":[115],"isolated":[120],"step,":[121],"FSLR":[122,149,167],"provides":[123],"explicit":[124],"supervision":[125],"unlike":[130],"CoT-SFT":[131,152],"implicitly":[133],"embeds":[134],"such":[135],"within":[137],"complete":[138],"solution":[139],"trajectories.Extensive":[140],"experiments":[141],"across":[142],"multiple":[143],"datasets":[146],"demonstrate":[147],"consistently":[150],"outperforms":[151],"under":[153],"both":[154],"in-distribution":[155],"out-of-distribution":[157],"settings,":[158],"average":[160],"improvements":[161],"3.2%":[163],"4.6%,":[165],"respectively.Moreover,":[166],"achieves":[168],"4-6\u00d7":[169],"faster":[170],"reduces":[173],"token":[175],"consumption":[176],"by":[177],"80%.":[179]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-07-02T00:00:00"}