{"id":"https://openalex.org/W7225596131","doi":"https://doi.org/10.5281/zenodo.24054273","title":"Prediction Is Not Output: One Head Cannot Do Both, Two Heads Can","display_name":"Prediction Is Not Output: One Head Cannot Do Both, Two Heads Can","publication_year":2026,"publication_date":"2026-11-08","ids":{"openalex":"https://openalex.org/W7225596131","doi":"https://doi.org/10.5281/zenodo.24054273"},"language":"en","primary_location":{"id":"doi:10.5281/zenodo.24054273","doi":"https://doi.org/10.5281/zenodo.24054273","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24054273","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.5281/zenodo.24054273","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5161302388","display_name":"Wenyuan Li","orcid":"https://orcid.org/0009-0004-2391-8508","observed_orcids":["https://orcid.org/0009-0004-2391-8508"]},"institutions":[{"id":"https://openalex.org/I118347636","display_name":"Australian National University","ror":"https://ror.org/019wvm592","country_code":"AU","type":"education","lineage":["https://openalex.org/I118347636"]}],"countries":["AU"],"is_corresponding":true,"raw_author_name":"Li, Wenyuan","raw_affiliation_strings":["Australian National University"],"raw_orcid":"https://orcid.org/0009-0004-2391-8508","affiliations":[{"raw_affiliation_string":"Australian National University","institution_ids":["https://openalex.org/I118347636"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5161302388"],"corresponding_institution_ids":["https://openalex.org/I118347636"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.7925000190734863,"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":0.7925000190734863,"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":0.0820000022649765,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.03750000149011612,"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/reinforcement-learning-from-human-feedback","display_name":"reinforcement learning from human feedback","score":0.8059999942779541},{"id":"https://openalex.org/keywords/supervised-fine-tuning","display_name":"supervised fine-tuning","score":0.847000002861023}],"concepts":[{"id":"https://openalex.org/C2780312720","wikidata":"https://www.wikidata.org/wiki/Q5689100","display_name":"Head (geology)","level":2,"score":0.745199978351593},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.645799994468689},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5674999952316284},{"id":"https://openalex.org/C90329073","wikidata":"https://www.wikidata.org/wiki/Q914232","display_name":"Ask price","level":2,"score":0.5471000075340271},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4894999861717224},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.4156999886035919},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.4083000123500824},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3921000063419342},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.36090001463890076},{"id":"https://openalex.org/C2777267654","wikidata":"https://www.wikidata.org/wiki/Q3519023","display_name":"Test (biology)","level":2,"score":0.33649998903274536},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3147999942302704},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2994000017642975},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.2833999991416931},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.27219998836517334},{"id":"https://openalex.org/C2994555788","wikidata":"https://www.wikidata.org/wiki/Q5689226","display_name":"Head start","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C169903167","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Test set","level":2,"score":0.2605000138282776}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.5281/zenodo.24054273","doi":"https://doi.org/10.5281/zenodo.24054273","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24054273","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},{"id":"doi:10.5281/zenodo.24054274","doi":"https://doi.org/10.5281/zenodo.24054274","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24054274","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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:10.5281/zenodo.24054273","doi":"https://doi.org/10.5281/zenodo.24054273","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24054273","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7223021041","https://openalex.org/W4401042580","https://openalex.org/W4415796289","https://openalex.org/W4389521044","https://openalex.org/W4404782729","https://openalex.org/W4385572248","https://openalex.org/W4312121705","https://openalex.org/W7213519439","https://openalex.org/W7220702684","https://openalex.org/W4400374467"],"abstract_inverted_index":{"A":[0,119,138,186],"language":[1,131],"model":[2,187],"uses":[3],"one":[4,61,195,200],"output":[5,74],"head":[6,62,121,140,175,228],"for":[7,31,86],"two":[8,28,81,189],"jobs:":[9],"predicting":[10],"what":[11],"other":[12,152,181],"people":[13,37],"write,":[14],"and":[15,52,126,199,233,245,248,259],"producing":[16],"its":[17,243,246],"own":[18],"replies.":[19],"Where":[20,35],"the":[21,27,32,36,39,80,89,159,173,178,180,225,237,240,250,255,257,260,271],"text":[22,40,261],"states":[23],"a":[24,46,53,76,107,111,192],"fact":[25],"correctly,":[26,148],"jobs":[29,66,144,206],"ask":[30],"same":[33,174],"output.":[34],"in":[38,110,208,221],"are":[41,91],"wrong,":[42],"they":[43],"do":[44,68],"not:":[45],"good":[47,54],"prediction":[48],"reproduces":[49],"their":[50],"mistakes,":[51],"reply":[55],"does":[56],"not.":[57],"We":[58,103],"prove":[59],"that":[60,218],"trained":[63,92,122,141,196,201],"on":[64,142,191,276],"both":[65,69,143,205],"cannot":[67],"well.":[70],"Its":[71],"best":[72],"possible":[73],"is":[75,84,117,169,215],"weighted":[77],"mixture":[78],"of":[79,158,164,235],"targets,":[82],"which":[83],"wrong":[85],"both,":[87],"whether":[88],"replies":[90],"by":[93,96,100,264],"supervised":[94],"fine-tuning,":[95],"reinforcement":[97],"learning":[98],"or":[99],"KL-regularized":[101],"RLHF.":[102],"test":[104],"this":[105],"with":[106,188],"small":[108],"transformer":[109],"synthetic":[112],"world":[113],"where":[114],"every":[115,209],"target":[116],"known.":[118],"single":[120,139],"first":[123],"to":[124,128,136,154,197,202],"predict":[125,198],"then":[127,150],"answer,":[129,203],"as":[130],"models":[132],"are,":[133],"forgets":[134],"how":[135],"predict.":[137],"at":[145],"once":[146],"answers":[147,177],"but":[149],"expects":[151],"speakers":[153,182],"answer":[155],"correctly":[156],"99%":[157],"time,":[160],"although":[161],"only":[162,220],"half":[163],"them":[165],"do;":[166],"when":[167],"answering":[168,227],"given":[170],"less":[171],"weight,":[172],"instead":[176],"way":[179],"do,":[183],"mistakes":[184],"included.":[185],"heads":[190],"shared":[193],"trunk,":[194],"keeps":[204],"right":[207],"setting":[210],"we":[211],"tried.":[212],"The":[213,252],"price":[214],"speed:":[216],"facts":[217],"appear":[219],"plain":[222],"statements":[223],"reach":[224],"separate":[226],"more":[229],"slowly.":[230],"Author":[231],"contributions":[232],"use":[234],"AI:":[236],"author":[238],"conceived":[239],"study,":[241],"set":[242],"question":[244],"claims,":[247],"directed":[249],"experiments.":[251],"experiment":[253],"code,":[254],"runs,":[256],"figures":[258],"were":[262],"produced":[263],"an":[265],"AI":[266],"assistant":[267],"(Claude,":[268],"Anthropic)":[269],"under":[270],"author's":[272],"direction.":[273],"Also":[274],"posted":[275],"aiXiv:":[277],"https://aixiv.science/abs/aixiv.261108.000002":[278]},"counts_by_year":[],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2026-11-09T00:00:00"}