{"id":"https://openalex.org/W7154505312","doi":"https://doi.org/10.48550/arxiv.2604.12426","title":"Do Transformers Use their Depth Adaptively? Evidence from a Relational Reasoning Task","display_name":"Do Transformers Use their Depth Adaptively? Evidence from a Relational Reasoning Task","publication_year":2026,"publication_date":"2026-04-14","ids":{"openalex":"https://openalex.org/W7154505312","doi":"https://doi.org/10.48550/arxiv.2604.12426"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.12426","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.12426","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":null,"license_id":null,"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.2604.12426","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133674881","display_name":"Alicia Curth","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Curth, Alicia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133661876","display_name":"Rachel Lawrence","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lawrence, Rachel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133714562","display_name":"Sushrut Karmalkar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Karmalkar, Sushrut","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5009328512","display_name":"Niranjani Prasad","orcid":"https://orcid.org/0000-0002-6677-280X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Prasad, Niranjani","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/T10656","display_name":"Child and Animal Learning Development","score":0.29820001125335693,"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"}},"topics":[{"id":"https://openalex.org/T10656","display_name":"Child and Animal Learning Development","score":0.29820001125335693,"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/T10465","display_name":"Neurobiology of Language and Bilingualism","score":0.15839999914169312,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T10730","display_name":"Language Development and Disorders","score":0.12839999794960022,"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/task","display_name":"Task (project management)","score":0.6363999843597412},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.566100001335144},{"id":"https://openalex.org/keywords/causal-chain","display_name":"Causal chain","score":0.4253999888896942},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4138999879360199},{"id":"https://openalex.org/keywords/multi-task-learning","display_name":"Multi-task learning","score":0.397599995136261},{"id":"https://openalex.org/keywords/causal-reasoning","display_name":"Causal reasoning","score":0.3928999900817871}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7896000146865845},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6363999843597412},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.59170001745224},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.566100001335144},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.43470001220703125},{"id":"https://openalex.org/C79897977","wikidata":"https://www.wikidata.org/wiki/Q5054568","display_name":"Causal chain","level":2,"score":0.4253999888896942},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.42289999127388},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4138999879360199},{"id":"https://openalex.org/C28006648","wikidata":"https://www.wikidata.org/wiki/Q6934509","display_name":"Multi-task learning","level":3,"score":0.397599995136261},{"id":"https://openalex.org/C115086926","wikidata":"https://www.wikidata.org/wiki/Q17004651","display_name":"Causal reasoning","level":3,"score":0.3928999900817871},{"id":"https://openalex.org/C169900460","wikidata":"https://www.wikidata.org/wiki/Q2200417","display_name":"Cognition","level":2,"score":0.361299991607666},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.35190001130104065},{"id":"https://openalex.org/C5655090","wikidata":"https://www.wikidata.org/wiki/Q192588","display_name":"Relational database","level":2,"score":0.2563000023365021},{"id":"https://openalex.org/C113336015","wikidata":"https://www.wikidata.org/wiki/Q574010","display_name":"Complete information","level":2,"score":0.2508000135421753}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.12426","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.12426","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.12426","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.12426","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.7594037055969238}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"We":[0],"investigate":[1],"whether":[2],"transformers":[3],"use":[4,93],"their":[5],"depth":[6,74,120],"adaptively":[7],"across":[8,44,59,99],"tasks":[9],"of":[10,31,118],"increasing":[11],"difficulty.":[12],"Using":[13],"a":[14],"controlled":[15],"multi-hop":[16],"relational":[17],"reasoning":[18],"task":[19],"based":[20],"on":[21,108],"family":[22],"stories,":[23],"where":[24],"difficulty":[25],"is":[26,57],"determined":[27],"by":[28],"the":[29,109,123],"number":[30],"relationship":[32],"hops":[33],"that":[34,132],"must":[35],"be":[36],"composed,":[37],"we":[38,67,111],"monitor":[39],"(i)":[40],"how":[41,54],"predictions":[42],"evolve":[43],"layers":[45,81,95],"via":[46,61],"early":[47],"readouts":[48],"(the":[49],"logit":[50],"lens)":[51],"and":[52,90,114],"(ii)":[53],"task-relevant":[55],"information":[56,98],"integrated":[58],"tokens":[60,100],"causal":[62],"patching.":[63],"For":[64,105],"pretrained":[65],"models,":[66],"find":[68,112],"some":[69,76],"limited":[70],"evidence":[71,117],"for":[72,87,127],"adaptive":[73,119],"use:":[75],"larger":[77],"models":[78,91,106],"need":[79],"fewer":[80],"to":[82,96],"arrive":[83],"at":[84],"plausible":[85],"answers":[86],"easier":[88],"tasks,":[89],"generally":[92],"more":[94,115],"integrate":[97],"as":[101],"chain":[102],"length":[103],"increases.":[104],"finetuned":[107],"task,":[110],"clearer":[113],"consistent":[116],"use,":[121],"with":[122],"effect":[124],"being":[125],"stronger":[126],"less":[128],"constrained":[129],"finetuning":[130],"regimes":[131],"do":[133],"not":[134],"preserve":[135],"general":[136],"language":[137],"modeling":[138],"abilities.":[139]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-16T00:00:00"}