{"id":"https://openalex.org/W7165886470","doi":"https://doi.org/10.48550/arxiv.2606.24898","title":"Dense Supervision Is Not Enough: The Readout Blind Spot in Looped Language Models","display_name":"Dense Supervision Is Not Enough: The Readout Blind Spot in Looped Language Models","publication_year":2026,"publication_date":"2026-06-12","ids":{"openalex":"https://openalex.org/W7165886470","doi":"https://doi.org/10.48550/arxiv.2606.24898"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.24898","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24898","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,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.24898","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133164990","display_name":"R Pavithra Sharma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sharma, Rituraj","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139314462","display_name":"Tu Vu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Vu, Tu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"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/T11948","display_name":"Machine Learning in Materials Science","score":0.18610000610351562,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11948","display_name":"Machine Learning in Materials Science","score":0.18610000610351562,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.10859999805688858,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.09459999948740005,"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/perplexity","display_name":"Perplexity","score":0.916700005531311},{"id":"https://openalex.org/keywords/usable","display_name":"USable","score":0.7857999801635742},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5647000074386597},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.5033000111579895},{"id":"https://openalex.org/keywords/residual","display_name":"Residual","score":0.4596000015735626},{"id":"https://openalex.org/keywords/variable","display_name":"Variable (mathematics)","score":0.4316999912261963},{"id":"https://openalex.org/keywords/state","display_name":"State (computer science)","score":0.3894999921321869},{"id":"https://openalex.org/keywords/train","display_name":"Train","score":0.3822000026702881}],"concepts":[{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.916700005531311},{"id":"https://openalex.org/C2780615836","wikidata":"https://www.wikidata.org/wiki/Q2471869","display_name":"USable","level":2,"score":0.7857999801635742},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6725999712944031},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5647000074386597},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.5033000111579895},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.4596000015735626},{"id":"https://openalex.org/C182365436","wikidata":"https://www.wikidata.org/wiki/Q50701","display_name":"Variable (mathematics)","level":2,"score":0.4316999912261963},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.39010000228881836},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.3894999921321869},{"id":"https://openalex.org/C190839683","wikidata":"https://www.wikidata.org/wiki/Q2448197","display_name":"Train","level":2,"score":0.3822000026702881},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.37599998712539673},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3222000002861023},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.3206999897956848},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.32030001282691956},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3052000105381012},{"id":"https://openalex.org/C94915269","wikidata":"https://www.wikidata.org/wiki/Q1834857","display_name":"Detector","level":2,"score":0.3034000098705292},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.30320000648498535},{"id":"https://openalex.org/C129537906","wikidata":"https://www.wikidata.org/wiki/Q7603913","display_name":"State variable","level":2,"score":0.299699991941452},{"id":"https://openalex.org/C73301696","wikidata":"https://www.wikidata.org/wiki/Q5469984","display_name":"Formalism (music)","level":3,"score":0.29919999837875366},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.298799991607666},{"id":"https://openalex.org/C2777067715","wikidata":"https://www.wikidata.org/wiki/Q3327726","display_name":"Multitaper","level":2,"score":0.29679998755455017},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.2863999903202057},{"id":"https://openalex.org/C26713055","wikidata":"https://www.wikidata.org/wiki/Q245962","display_name":"Implementation","level":2,"score":0.2766999900341034}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.24898","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24898","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,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.24898","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24898","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,"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"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","score":0.7253865003585815,"id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Looped":[0],"language":[1],"models":[2],"turn":[3],"hidden":[4],"states":[5],"into":[6,18,120],"runtime":[7],"state:":[8],"each":[9],"state":[10,28],"is":[11,141,150],"decoded":[12],"for":[13],"prediction":[14],"and":[15,67,84,103,129,138],"fed":[16],"back":[17],"future":[19],"computation.":[20],"This":[21],"creates":[22],"a":[23,58,165],"basic":[24],"supervision":[25,153],"question:":[26],"which":[27],"variables":[29,42],"does":[30],"cross-entropy":[31,39,75,111],"actually":[32],"control?":[33],"We":[34],"show":[35],"that":[36,86],"dense":[37,152],"per-loop":[38,90,110],"controls":[40],"the":[41,45,52,73,121,136,142,171],"exposed":[43],"by":[44],"readout,":[46],"not":[47],"every":[48],"variable":[49],"active":[50],"in":[51,135,187],"recurrent":[53,99,156],"transition.":[54],"Hidden-state":[55],"scale":[56,71,157,162],"gives":[57],"concrete":[59],"failure":[60],"mode.":[61],"Scale-invariant":[62],"readouts":[63,114,128],"such":[64],"as":[65],"RMSNorm":[66,113],"LayerNorm":[68],"hide":[69],"radial":[70],"from":[72,170],"immediate":[74],"loss,":[76],"while":[77],"pre-norm":[78],"residual":[79],"recurrence":[80,140],"continues":[81],"to":[82,164],"carry":[83],"update":[85],"same":[87],"scale.":[88,100],"Thus":[89],"loss":[91,166],"can":[92],"make":[93],"early":[94],"exits":[95],"usable":[96],"without":[97,107],"controlling":[98],"In":[101],"44M":[102],"129M":[104],"looped":[105],"transformers":[106],"inter-loop":[108],"normalization,":[109],"through":[112],"still":[115],"drives":[116],"final":[117],"hidden-state":[118],"norms":[119,134],"thousands":[122],"or":[123,167],"tens":[124],"of":[125],"thousands.":[126],"Scale-visible":[127],"explicit":[130],"norm":[131],"penalties":[132],"keep":[133],"tens,":[137],"scale-removing":[139],"complementary":[143],"architectural":[144],"fix.":[145],"The":[146],"resulting":[147],"design":[148],"rule":[149],"simple:":[151],"trains":[154],"exits;":[155],"control":[158],"requires":[159],"either":[160],"making":[161],"visible":[163],"removing":[168],"it":[169],"loop.":[172],"Consistent":[173],"with":[174],"this":[175],"rule,":[176],"scale-controlled":[177],"variants":[178],"achieve":[179],"lower":[180],"perplexity":[181],"at":[182],"matched":[183],"inference-depth":[184],"operating":[185],"points":[186],"our":[188],"variable-depth":[189],"benchmarks.":[190]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-06-26T00:00:00"}