{"id":"https://openalex.org/W4416036331","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.904","title":"Answer Convergence as a Signal for Early Stopping in Reasoning","display_name":"Answer Convergence as a Signal for Early Stopping in Reasoning","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416036331","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.904"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.904","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.904","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.904","pdf_url":"https://aclanthology.org/2025.emnlp-main.904.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","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":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2025.emnlp-main.904.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100737898","display_name":"Xin Liu","orcid":"https://orcid.org/0000-0003-3051-4793","observed_orcids":["https://orcid.org/0000-0003-3051-4793","https://orcid.org/0000-0002-0440-3638"]},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Xin Liu","raw_affiliation_strings":["Computer Science and Engineering University of Michigan Ann Arbor , MI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Computer Science and Engineering University of Michigan Ann Arbor , MI","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100364452","display_name":"Lu Wang","orcid":"https://orcid.org/0000-0002-4614-4427","observed_orcids":["https://orcid.org/0000-0002-4614-4427"]},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Lu Wang","raw_affiliation_strings":["Computer Science and Engineering University of Michigan Ann Arbor , MI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Computer Science and Engineering University of Michigan Ann Arbor , MI","institution_ids":["https://openalex.org/I27837315"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I27837315"],"apc_list":null,"apc_paid":null,"fwci":0.9483,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.77247436,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":94},"biblio":{"volume":null,"issue":null,"first_page":"17907","last_page":"17918"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.8715999722480774,"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":0.8715999722480774,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.0940999984741211,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.010300000198185444,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/chain-of-thought-prompting","display_name":"chain-of-thought prompting","score":0.9120000004768372},{"id":"https://openalex.org/keywords/early-stopping","display_name":"early stopping","score":0.9079999923706055},{"id":"https://openalex.org/keywords/answer-consistency","display_name":"answer consistency","score":0.7390000224113464},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.6980000138282776},{"id":"https://openalex.org/keywords/reasoning-efficiency","display_name":"reasoning efficiency","score":0.625},{"id":"https://openalex.org/keywords/token-usage-reduction","display_name":"token usage reduction","score":0.8669999837875366},{"id":"https://openalex.org/keywords/inference-cost","display_name":"inference cost","score":0.6240000128746033},{"id":"https://openalex.org/keywords/internal-activations","display_name":"internal activations","score":0.6899999976158142},{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.675000011920929},{"id":"https://openalex.org/keywords/open-weight-llms","display_name":"open-weight LLMs","score":0.6290000081062317}],"concepts":[{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.5455999970436096},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5052000284194946},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.4973999857902527},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4318999946117401},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.414000004529953},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3882000148296356},{"id":"https://openalex.org/C104267543","wikidata":"https://www.wikidata.org/wiki/Q208163","display_name":"Signal processing","level":3,"score":0.3221000134944916},{"id":"https://openalex.org/C2777686260","wikidata":"https://www.wikidata.org/wiki/Q144037","display_name":"Calculus (dental)","level":2,"score":0.3192000091075897},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.3001999855041504},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2768999934196472},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.27390000224113464},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2531000077724457},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.2502000033855438}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.904","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.904","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.904","pdf_url":"https://aclanthology.org/2025.emnlp-main.904.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","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":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2506.02536","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2506.02536","pdf_url":"https://arxiv.org/pdf/2506.02536","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2506.02536","doi":"https://doi.org/10.48550/arxiv.2506.02536","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2506.02536","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.904","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.904","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.904","pdf_url":"https://aclanthology.org/2025.emnlp-main.904.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","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":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G4230946465","display_name":"CAREER: Long Document Summarization with Question-Summary Hierarchy and User Preference Control","funder_award_id":"2046016","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320313434","display_name":"Institute for Catastrophic Loss Reduction","ror":"https://ror.org/03prrhh04"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416036331.pdf","grobid_xml":"https://content.openalex.org/works/W4416036331.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4416255603","https://openalex.org/W4415329263","https://openalex.org/W7128616912","https://openalex.org/W7161709774","https://openalex.org/W4417226708","https://openalex.org/W7168030527","https://openalex.org/W7168033419","https://openalex.org/W7168031077","https://openalex.org/W4417099852","https://openalex.org/W7156370138"],"abstract_inverted_index":{"Chain-of-thought":[0],"(CoT)":[1],"prompting":[2],"enhances":[3],"reasoning":[4,25,47,60,75,164],"in":[5,80],"large":[6],"language":[7],"models":[8,64],"(LLMs)":[9],"but":[10],"often":[11],"leads":[12],"to":[13,41,52,67,92,117],"verbose":[14],"and":[15,109,126],"redundant":[16],"outputs,":[17],"thus":[18],"increasing":[19],"inference":[20,169],"cost.We":[21],"hypothesize":[22],"that":[23,58,114,131,166],"many":[24],"steps":[26],"are":[27],"unnecessary":[28],"for":[29,49,174],"producing":[30],"correct":[31],"answers.To":[32],"investigate":[33],"this,":[34],"we":[35,87],"start":[36],"with":[37,138],"a":[38,50,54,111],"systematic":[39],"study":[40],"examine":[42],"what":[43],"is":[44],"the":[45,74,81,103,160],"minimum":[46],"required":[48],"model":[51],"reach":[53],"stable":[55],"decision.We":[56],"find":[57],"on":[59,84,120,145],"tasks":[61],"like":[62],"math,":[63],"typically":[65],"converge":[66],"their":[68],"final":[69],"answers":[70],"after":[71],"60%":[72],"of":[73,105,162],"steps,":[76],"suggesting":[77],"substantial":[78],"redundancy":[79],"remaining":[82],"content.Based":[83],"these":[85],"insights,":[86],"propose":[88],"three":[89],"inference-time":[90],"strategies":[91],"improve":[93],"efficiency:":[94],"(1)":[95],"early":[96],"stopping":[97],"via":[98],"answer":[99],"consistency,":[100],"(2)":[101],"boosting":[102],"probability":[104],"generating":[106],"end-of-reasoning":[107],"signals,":[108],"(3)":[110],"supervised":[112],"method":[113],"learns":[115],"when":[116],"stop":[118],"based":[119],"internal":[121],"activations.Experiments":[122],"across":[123],"five":[124,127],"benchmarks":[125],"open-weights":[128],"LLMs":[129],"show":[130],"our":[132],"methods":[133,165],"significantly":[134],"reduce":[135],"token":[136],"usage":[137],"little":[139],"or":[140],"no":[141],"accuracy":[142],"drop.In":[143],"particular,":[144],"NaturalQuestions,":[146],"Answer":[147],"Consistency":[148],"reduces":[149],"tokens":[150],"by":[151],"over":[152],"40%":[153],"while":[154],"further":[155],"improving":[156],"accuracy.Our":[157],"work":[158],"underscores":[159],"importance":[161],"cost-effective":[163],"operate":[167],"at":[168],"time,":[170],"offering":[171],"practical":[172],"benefits":[173],"real-world":[175],"applications.1":[176]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-10-15T10:00:46.086208","created_date":"2025-11-08T00:00:00"}