{"id":"https://openalex.org/W7166833167","doi":"https://doi.org/10.18653/v1/2026.findings-acl.622","title":"Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning","display_name":"Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166833167","doi":"https://doi.org/10.18653/v1/2026.findings-acl.622"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.622","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.622","pdf_url":"https://aclanthology.org/2026.findings-acl.622.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 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.622.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139744516","display_name":"Nathana\u00ebl Carraz Rakotonirina","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nathana\u00ebl Carraz Rakotonirina","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021575600","display_name":"Ren Pang","orcid":"https://orcid.org/0009-0000-2474-4491"},"institutions":[{"id":"https://openalex.org/I170486558","display_name":"Universitat Pompeu Fabra","ror":"https://ror.org/04n0g0b29","country_code":"ES","type":"education","lineage":["https://openalex.org/I170486558"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Ren Pang","raw_affiliation_strings":["Universitat Pompeu Fabra \u2663 AWS AI Labs \u2660 Amazon AGI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Universitat Pompeu Fabra \u2663 AWS AI Labs \u2660 Amazon AGI","institution_ids":["https://openalex.org/I170486558"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5010761363","display_name":"Neha Anna John","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Neha Anna John","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139723056","display_name":"Michael Bohlke-Schneider","orcid":null},"institutions":[{"id":"https://openalex.org/I170486558","display_name":"Universitat Pompeu Fabra","ror":"https://ror.org/04n0g0b29","country_code":"ES","type":"education","lineage":["https://openalex.org/I170486558"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Michael Bohlke-Schneider","raw_affiliation_strings":["Universitat Pompeu Fabra \u2663 AWS AI Labs \u2660 Amazon AGI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Universitat Pompeu Fabra \u2663 AWS AI Labs \u2660 Amazon AGI","institution_ids":["https://openalex.org/I170486558"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5076120409","display_name":"Momchil Hardalov","orcid":"https://orcid.org/0000-0001-8095-3570"},"institutions":[{"id":"https://openalex.org/I170486558","display_name":"Universitat Pompeu Fabra","ror":"https://ror.org/04n0g0b29","country_code":"ES","type":"education","lineage":["https://openalex.org/I170486558"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Momchil Hardalov","raw_affiliation_strings":["Universitat Pompeu Fabra \u2663 AWS AI Labs \u2660 Amazon AGI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Universitat Pompeu Fabra \u2663 AWS AI Labs \u2660 Amazon AGI","institution_ids":["https://openalex.org/I170486558"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.82288009,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"12798","last_page":"12810"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11902","display_name":"Intelligent Tutoring Systems and Adaptive Learning","score":0.26840001344680786,"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/T11902","display_name":"Intelligent Tutoring Systems and Adaptive Learning","score":0.26840001344680786,"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/T13447","display_name":"Educational and Psychological Assessments","score":0.05730000138282776,"subfield":{"id":"https://openalex.org/subfields/3204","display_name":"Developmental and Educational Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10636","display_name":"Innovative Teaching and Learning Methods","score":0.05689999833703041,"subfield":{"id":"https://openalex.org/subfields/3204","display_name":"Developmental and Educational Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.5328999757766724},{"id":"https://openalex.org/keywords/case-based-reasoning","display_name":"Case-based reasoning","score":0.28700000047683716},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.2542000114917755},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.2533000111579895},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.24770000576972961}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6032000184059143},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.5328999757766724},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5235999822616577},{"id":"https://openalex.org/C20162079","wikidata":"https://www.wikidata.org/wiki/Q1151406","display_name":"Case-based reasoning","level":2,"score":0.28700000047683716},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.2542000114917755},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.2533000111579895},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.24770000576972961},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.24369999766349792},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.24220000207424164},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.2386000007390976}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.622","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.622","pdf_url":"https://aclanthology.org/2026.findings-acl.622.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 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.622","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.622","pdf_url":"https://aclanthology.org/2026.findings-acl.622.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 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320320300","display_name":"European Commission","ror":"https://ror.org/00k4n6c32"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166833167.pdf","grobid_xml":"https://content.openalex.org/works/W7166833167.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"reasoning":[1,51,60,100,149],"capabilities":[2],"of":[3,19,114,130],"large":[4],"language":[5],"models":[6,118],"(LLMs)":[7],"have":[8],"improved":[9],"substantially":[10],"through":[11],"increased":[12],"test-time":[13],"computation,":[14],"typically":[15],"in":[16],"the":[17,80,85,103,154,157,166,173],"form":[18],"intermediate":[20],"tokens":[21,77],"known":[22,44],"as":[23,45],"chain-of-thought":[24],"(CoT).However,":[25],"CoT":[26],"often":[27],"becomes":[28],"unnecessarily":[29],"long,":[30],"increasing":[31],"computation":[32],"costs":[33],"without":[34],"improving":[35],"accuracy":[36],"and":[37,119,132,169],"sometimes":[38],"even":[39],"degrading":[40],"performance,":[41],"a":[42,48,71,94,141],"phenomenon":[43],"\"overthinking\".We":[46],"propose":[47],"multi-stage":[49],"efficient":[50,148],"method":[52],"that":[53,75],"combines":[54],"supervised":[55],"finetuning-via":[56],"rejection":[57],"sampling":[58],"or":[59],"trace":[61],"reformatting-with":[62],"reinforcement":[63],"learning":[64],"using":[65],"an":[66,112],"adaptive":[67],"length":[68,105,110],"penalty.We":[69],"introduce":[70],"lightweight":[72],"reward":[73],"function":[74],"penalizes":[76],"generated":[78],"after":[79],"first":[81],"correct":[82],"answer,":[83],"encouraging":[84],"model":[86,168],"to":[87],"perform":[88],"selfverification":[89],"only":[90,126],"when":[91],"beneficial.We":[92],"conduct":[93],"holistic":[95],"evaluation":[96],"across":[97],"seven":[98],"diverse":[99],"tasks,":[101],"analyzing":[102],"accuracy-response":[104],"trade-off.Our":[106],"approach":[107],"reduces":[108],"response":[109],"by":[111],"average":[113],"28%":[115],"for":[116,121],"8B":[117],"40%":[120],"32B":[122],"models,":[123],"while":[124],"incurring":[125],"minor":[127],"performance":[128],"drops":[129],"1.6":[131],"2.5":[133,170],"points,":[134],"respectively.Despite":[135],"its":[136],"conceptual":[137],"simplicity,":[138],"it":[139],"achieves":[140],"better":[142],"trade-off":[143],"than":[144],"more":[145],"complex":[146],"state-of-the-art":[147],"methods,":[150],"scoring":[151],"76.6":[152],"on":[153],"area":[155],"under":[156],"Overthinking-Adjusted":[158],"Accuracy":[159],"curve":[160],"(AUC":[161],"OAA":[162],")-5":[163],"points":[164,171],"above":[165,172],"base":[167],"second-best":[174],"approach.":[175]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}