{"id":"https://openalex.org/W4412888294","doi":"https://doi.org/10.18653/v1/2025.findings-acl.662","title":"Knowing Before Saying: LLM Representations Encode Information About Chain-of-Thought Success Before Completion","display_name":"Knowing Before Saying: LLM Representations Encode Information About Chain-of-Thought Success Before Completion","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888294","doi":"https://doi.org/10.18653/v1/2025.findings-acl.662"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.662","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.662","pdf_url":"https://aclanthology.org/2025.findings-acl.662.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.662.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5085291599","display_name":"Anum Afzal","orcid":"https://orcid.org/0000-0002-3081-7914"},"institutions":[{"id":"https://openalex.org/I1304085615","display_name":"Nvidia (United Kingdom)","ror":"https://ror.org/02kr42612","country_code":"GB","type":"company","lineage":["https://openalex.org/I1304085615","https://openalex.org/I4210127875"]},{"id":"https://openalex.org/I62916508","display_name":"Technical University of Munich","ror":"https://ror.org/02kkvpp62","country_code":"DE","type":"education","lineage":["https://openalex.org/I62916508"]}],"countries":["DE","GB"],"is_corresponding":false,"raw_author_name":"Anum Afzal","raw_affiliation_strings":["Technical University of Munich Technical University of Munich Nvidia Research & Bar-Ilan University Nvidia Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technical University of Munich Technical University of Munich Nvidia Research & Bar-Ilan University Nvidia Research","institution_ids":["https://openalex.org/I1304085615","https://openalex.org/I62916508"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022973212","display_name":"Florian Matthes","orcid":"https://orcid.org/0000-0002-6667-5452"},"institutions":[{"id":"https://openalex.org/I1304085615","display_name":"Nvidia (United Kingdom)","ror":"https://ror.org/02kr42612","country_code":"GB","type":"company","lineage":["https://openalex.org/I1304085615","https://openalex.org/I4210127875"]},{"id":"https://openalex.org/I62916508","display_name":"Technical University of Munich","ror":"https://ror.org/02kkvpp62","country_code":"DE","type":"education","lineage":["https://openalex.org/I62916508"]}],"countries":["DE","GB"],"is_corresponding":false,"raw_author_name":"Florian Matthes","raw_affiliation_strings":["Technical University of Munich Technical University of Munich Nvidia Research & Bar-Ilan University Nvidia Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technical University of Munich Technical University of Munich Nvidia Research & Bar-Ilan University Nvidia Research","institution_ids":["https://openalex.org/I1304085615","https://openalex.org/I62916508"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045719865","display_name":"Gal Chechik","orcid":"https://orcid.org/0000-0001-9164-5303"},"institutions":[{"id":"https://openalex.org/I1304085615","display_name":"Nvidia (United Kingdom)","ror":"https://ror.org/02kr42612","country_code":"GB","type":"company","lineage":["https://openalex.org/I1304085615","https://openalex.org/I4210127875"]},{"id":"https://openalex.org/I62916508","display_name":"Technical University of Munich","ror":"https://ror.org/02kkvpp62","country_code":"DE","type":"education","lineage":["https://openalex.org/I62916508"]}],"countries":["DE","GB"],"is_corresponding":false,"raw_author_name":"Gal Chechik","raw_affiliation_strings":["Technical University of Munich Technical University of Munich Nvidia Research & Bar-Ilan University Nvidia Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technical University of Munich Technical University of Munich Nvidia Research & Bar-Ilan University Nvidia Research","institution_ids":["https://openalex.org/I1304085615","https://openalex.org/I62916508"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5003744048","display_name":"Yftah Ziser","orcid":null},"institutions":[{"id":"https://openalex.org/I1304085615","display_name":"Nvidia (United Kingdom)","ror":"https://ror.org/02kr42612","country_code":"GB","type":"company","lineage":["https://openalex.org/I1304085615","https://openalex.org/I4210127875"]},{"id":"https://openalex.org/I62916508","display_name":"Technical University of Munich","ror":"https://ror.org/02kkvpp62","country_code":"DE","type":"education","lineage":["https://openalex.org/I62916508"]}],"countries":["DE","GB"],"is_corresponding":false,"raw_author_name":"Yftah Ziser","raw_affiliation_strings":["Technical University of Munich Technical University of Munich Nvidia Research & Bar-Ilan University Nvidia Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technical University of Munich Technical University of Munich Nvidia Research & Bar-Ilan University Nvidia Research","institution_ids":["https://openalex.org/I1304085615","https://openalex.org/I62916508"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"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":0.14338346,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"12791","last_page":"12806"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T13083","display_name":"Advanced Text Analysis Techniques","score":0.8858000040054321,"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/T13083","display_name":"Advanced Text Analysis Techniques","score":0.8858000040054321,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.8801000118255615,"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":0.873199999332428,"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/encode","display_name":"ENCODE","score":0.7673125267028809},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6438313722610474},{"id":"https://openalex.org/keywords/cognitive-science","display_name":"Cognitive science","score":0.44148606061935425},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.406684547662735},{"id":"https://openalex.org/keywords/psychology","display_name":"Psychology","score":0.2594822943210602},{"id":"https://openalex.org/keywords/gene","display_name":"Gene","score":0.07226875424385071},{"id":"https://openalex.org/keywords/chemistry","display_name":"Chemistry","score":0.06551042199134827}],"concepts":[{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.7673125267028809},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6438313722610474},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.44148606061935425},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.406684547662735},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.2594822943210602},{"id":"https://openalex.org/C104317684","wikidata":"https://www.wikidata.org/wiki/Q7187","display_name":"Gene","level":2,"score":0.07226875424385071},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.06551042199134827},{"id":"https://openalex.org/C55493867","wikidata":"https://www.wikidata.org/wiki/Q7094","display_name":"Biochemistry","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.662","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.662","pdf_url":"https://aclanthology.org/2025.findings-acl.662.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"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.662","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.662","pdf_url":"https://aclanthology.org/2025.findings-acl.662.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":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888294.pdf","grobid_xml":"https://content.openalex.org/works/W4412888294.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2468279273","https://openalex.org/W2354198838","https://openalex.org/W1989130879","https://openalex.org/W2103419012","https://openalex.org/W1974414866","https://openalex.org/W2057568687","https://openalex.org/W2063982682"],"abstract_inverted_index":{"We":[0],"investigate":[1],"whether":[2],"the":[3,39,46,59],"success":[4],"of":[5],"a":[6,18,29,51,131],"zeroshot":[7],"Chain-of-Thought":[8],"(CoT)":[9],"process":[10,41],"can":[11,103],"be":[12],"predicted":[13],"before":[14,28],"completion.We":[15],"discover":[16],"that":[17,35,117,165],"probing":[19],"classifier,":[20],"based":[21],"on":[22,58,67],"LLM":[23],"representations,":[24],"performs":[25,62],"well":[26],"even":[27],"single":[30],"token":[31],"is":[32,42,87,160],"generated,":[33],"suggesting":[34,95],"crucial":[36],"information":[37,99],"about":[38],"reasoning":[40,74,78,102,120],"already":[43],"present":[44],"in":[45],"initial":[47],"steps":[48,79],"representations.In":[49],"contrast,":[50],"strong":[52],"BERT-based":[53],"baseline,":[54],"which":[55],"relies":[56],"solely":[57],"generated":[60],"tokens,":[61],"worse-likely":[63],"because":[64],"it":[65],"depends":[66],"shallow":[68],"linguistic":[69],"cues":[70],"rather":[71],"than":[72],"deeper":[73],"dynamics.Surprisingly,":[75],"using":[76,126],"later":[77,92],"does":[80],"not":[81,125],"always":[82],"improve":[83],"classification.When":[84],"additional":[85],"context":[86],"unhelpful,":[88],"earlier":[89],"representations":[90],"resemble":[91],"ones":[93],"more,":[94],"LLMs":[96],"encode":[97],"key":[98],"early.This":[100],"implies":[101],"often":[104],"stop":[105],"early":[106,113,158],"without":[107],"loss.To":[108],"test":[109],"this,":[110],"we":[111],"conduct":[112],"stopping":[114,159],"experiments,":[115],"showing":[116],"truncating":[118],"CoT":[119,127,148],"still":[121],"improves":[122],"performance":[123],"over":[124],"at":[128],"all,":[129],"though":[130],"gap":[132],"remains":[133],"compared":[134],"to":[135,146,155,171],"full":[136],"reasoning.However,":[137],"approaches":[138],"like":[139],"supervised":[140],"learning":[141,144],"or":[142],"reinforcement":[143],"designed":[145],"shorten":[147],"chains":[149],"could":[150],"leverage":[151],"our":[152],"classifier's":[153],"guidance":[154],"identify":[156],"when":[157],"effective.Our":[161],"findings":[162],"provide":[163],"insights":[164],"may":[166],"support":[167],"such":[168],"methods,":[169],"helping":[170],"optimize":[172],"CoT's":[173],"efficiency":[174],"while":[175],"preserving":[176],"its":[177],"benefits.":[178]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}