{"id":"https://openalex.org/W7166880660","doi":"https://doi.org/10.18653/v1/2026.acl-long.60","title":"Crosscoding Through Time: Tracking Emergence & Consolidation Of Linguistic Representations Throughout LLM Pretraining","display_name":"Crosscoding Through Time: Tracking Emergence & Consolidation Of Linguistic Representations Throughout LLM Pretraining","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166880660","doi":"https://doi.org/10.18653/v1/2026.acl-long.60"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.acl-long.60","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.60","pdf_url":"https://aclanthology.org/2026.acl-long.60.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.acl-long.60.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5018905679","display_name":"Deniz Bayazit","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deniz Bayazit","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5020998070","display_name":"Aaron Mueller","orcid":"https://orcid.org/0009-0005-1148-5001"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Aaron Mueller","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139772678","display_name":"Antoine Bosselut","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Antoine Bosselut","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.86905522,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1353","last_page":"1377"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.40790000557899475,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.40790000557899475,"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/T11587","display_name":"Second Language Acquisition and Learning","score":0.10909999907016754,"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/T12090","display_name":"Language and cultural evolution","score":0.10010000318288803,"subfield":{"id":"https://openalex.org/subfields/3316","display_name":"Cultural Studies"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/consolidation","display_name":"Consolidation (business)","score":0.44679999351501465},{"id":"https://openalex.org/keywords/tracking","display_name":"Tracking (education)","score":0.37470000982284546},{"id":"https://openalex.org/keywords/eye-tracking","display_name":"Eye tracking","score":0.2563999891281128}],"concepts":[{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.5777000188827515},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.4620000123977661},{"id":"https://openalex.org/C2776014549","wikidata":"https://www.wikidata.org/wiki/Q3050847","display_name":"Consolidation (business)","level":2,"score":0.44679999351501465},{"id":"https://openalex.org/C2775936607","wikidata":"https://www.wikidata.org/wiki/Q466845","display_name":"Tracking (education)","level":2,"score":0.37470000982284546},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.35109999775886536},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.34220001101493835},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.34040001034736633},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.33869999647140503},{"id":"https://openalex.org/C46312422","wikidata":"https://www.wikidata.org/wiki/Q11024","display_name":"Communication","level":1,"score":0.3278000056743622},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.26010000705718994},{"id":"https://openalex.org/C56461940","wikidata":"https://www.wikidata.org/wiki/Q970687","display_name":"Eye tracking","level":2,"score":0.2563999891281128}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.60","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.60","pdf_url":"https://aclanthology.org/2026.acl-long.60.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.60","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.60","pdf_url":"https://aclanthology.org/2026.acl-long.60.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3056497733","display_name":"Interfacing Neural and Symbolic Knowledge in Natural Language Systems","funder_award_id":"215390","funder_id":"https://openalex.org/F4320320924","funder_display_name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320320924","display_name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung","ror":"https://ror.org/00yjd3n13"},{"id":"https://openalex.org/F4320337144","display_name":"National Supercomputing Center, Korea Institute of Science and Technology Information","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166880660.pdf","grobid_xml":"https://content.openalex.org/works/W7166880660.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,26],"(LLMs)":[3],"learn":[4],"nontrivial":[5],"abstractions":[6],"during":[7,78,125],"pretraining,":[8],"such":[9],"as":[10],"detecting":[11],"irregular":[12],"plural":[13],"noun":[14],"subjects.However,":[15],"because":[16],"traditional":[17],"evaluation":[18],"methods":[19],"(e.g.,":[20],"benchmarking)":[21],"fail":[22],"to":[23,61,101],"reveal":[24],"how":[25,39],"acquire":[27],"these":[28,40],"concepts":[29],"and":[30,38,48,63,89,92,123,130,139],"capabilities,":[31],"it":[32],"is":[33,128],"not":[34],"well":[35],"understood":[36],"when":[37],"specific":[41],"linguistic":[42,76],"abilities":[43],"emerge.To":[44],"bridge":[45],"this":[46,69],"gap":[47],"better":[49],"understand":[50],"model":[51,67],"training":[52,103],"at":[53,105],"the":[54,73],"concept":[55],"level,":[56],"we":[57,71],"use":[58],"sparse":[59],"crosscoders":[60,81,117],"discover":[62],"align":[64],"features":[65,77,108],"across":[66],"checkpoints.Using":[68],"approach,":[70],"track":[72],"evolution":[74],"of":[75,142],"pretraining.We":[79],"train":[80],"between":[82],"opensourced":[83],"checkpoint":[84],"triplets":[85],"with":[86],"significant":[87],"performance":[88],"representation":[90,143],"shifts,":[91],"introduce":[93],"a":[94,133],"novel":[95],"metric,":[96],"Relative":[97],"Indirect":[98],"Effects":[99],"(RELIE),":[100],"trace":[102],"stages":[104],"which":[106],"individual":[107],"become":[109],"causally":[110],"important":[111],"for":[112],"task":[113],"performance.We":[114],"show":[115],"that":[116],"can":[118],"detect":[119],"feature":[120],"emergence,":[121],"maintenance,":[122],"discontinuation":[124],"pretraining.Our":[126],"approach":[127],"architecture-agnostic":[129],"scalable,":[131],"offering":[132],"promising":[134],"path":[135],"toward":[136],"more":[137],"interpretable":[138],"fine-grained":[140],"analysis":[141],"learning":[144],"throughout":[145],"pretraining.":[146],"1":[147]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}