{"id":"https://openalex.org/W7204448133","doi":"https://doi.org/10.18653/2026.emnlp-main.701","title":"The Dialect Tax: Dialectal Biases Persist throughout the Language Modeling Pipeline","display_name":"The Dialect Tax: Dialectal Biases Persist throughout the Language Modeling Pipeline","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7204448133","doi":"https://doi.org/10.18653/2026.emnlp-main.701"},"language":"en","primary_location":{"id":"doi:10.18653/2026.emnlp-main.701","doi":"https://doi.org/10.18653/2026.emnlp-main.701","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.701","pdf_url":"https://aclanthology.org/2026.emnlp-main.701.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2026.emnlp-main.701.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5148449590","display_name":"Elle","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Elle","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5148449590"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.68273779,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"16184","last_page":"16206"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.5214999914169312,"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":0.5214999914169312,"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":0.19850000739097595,"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/T11640","display_name":"Linguistic Variation and Morphology","score":0.10670000314712524,"subfield":{"id":"https://openalex.org/subfields/3310","display_name":"Linguistics and Language"},"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/language-models","display_name":"language models","score":0.6710000038146973},{"id":"https://openalex.org/keywords/african-american-english","display_name":"African American English","score":0.75},{"id":"https://openalex.org/keywords/standard-american-english","display_name":"Standard American English","score":0.8579999804496765},{"id":"https://openalex.org/keywords/tokenization","display_name":"tokenization","score":0.6620000004768372},{"id":"https://openalex.org/keywords/subword-tokenization","display_name":"subword tokenization","score":0.7269999980926514},{"id":"https://openalex.org/keywords/character-level-tokenization","display_name":"character-level tokenization","score":0.7749999761581421},{"id":"https://openalex.org/keywords/pretraining","display_name":"pre-training","score":0.628000020980835},{"id":"https://openalex.org/keywords/post-training","display_name":"post-training","score":0.718999981880188},{"id":"https://openalex.org/keywords/reward-models","display_name":"reward models","score":0.875}],"concepts":[{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.7537999749183655},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6789000034332275},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.656000018119812},{"id":"https://openalex.org/C108650721","wikidata":"https://www.wikidata.org/wiki/Q1783253","display_name":"Counterfactual thinking","level":2,"score":0.6510999798774719},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.5703999996185303},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5695000290870667},{"id":"https://openalex.org/C2780876879","wikidata":"https://www.wikidata.org/wiki/Q3054749","display_name":"Meaning (existential)","level":2,"score":0.5257999897003174},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5054000020027161},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.4027000069618225},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.3894999921321869},{"id":"https://openalex.org/C89600930","wikidata":"https://www.wikidata.org/wiki/Q1423946","display_name":"Segmentation","level":2,"score":0.34529998898506165},{"id":"https://openalex.org/C2777939226","wikidata":"https://www.wikidata.org/wiki/Q7976","display_name":"American English","level":2,"score":0.3089999854564667},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.30730000138282776},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.28870001435279846},{"id":"https://openalex.org/C2776608160","wikidata":"https://www.wikidata.org/wiki/Q4785462","display_name":"Natural (archaeology)","level":2,"score":0.28679999709129333},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2831999957561493},{"id":"https://openalex.org/C155092808","wikidata":"https://www.wikidata.org/wiki/Q182557","display_name":"Computational linguistics","level":2,"score":0.2831000089645386},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.271699994802475},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.25209999084472656},{"id":"https://openalex.org/C171041071","wikidata":"https://www.wikidata.org/wiki/Q36870","display_name":"First language","level":2,"score":0.250900000333786}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/2026.emnlp-main.701","doi":"https://doi.org/10.18653/2026.emnlp-main.701","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.701","pdf_url":"https://aclanthology.org/2026.emnlp-main.701.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2608.24952","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24952","pdf_url":"https://arxiv.org/pdf/2608.24952","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.24952","doi":"https://doi.org/10.48550/arxiv.2608.24952","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.24952","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,"listed_in":[],"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"}],"best_oa_location":{"id":"doi:10.18653/2026.emnlp-main.701","doi":"https://doi.org/10.18653/2026.emnlp-main.701","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.701","pdf_url":"https://aclanthology.org/2026.emnlp-main.701.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/10","display_name":"Reduced inequalities","score":0.7578999996185303}],"sustainable_development_goals_aurora":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality education","score":0.6049107909202576}],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204448133.pdf","grobid_xml":"https://content.openalex.org/works/W7204448133.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4416035598","https://openalex.org/W7163137410","https://openalex.org/W7164130904","https://openalex.org/W7214044436","https://openalex.org/W7123345850","https://openalex.org/W4416034421","https://openalex.org/W4416034762","https://openalex.org/W7167826577","https://openalex.org/W7167586886","https://openalex.org/W7167784251"],"abstract_inverted_index":{"Systematic":[0],"dialectal":[1,60,83,108,132],"performance":[2,73],"gaps":[3,69],"in":[4,186],"language":[5,20,33,194],"models":[6,128,144],"(LMs)":[7],"are":[8],"well":[9],"documented,":[10],"but":[11,188],"the":[12,18,31,174,193],"source":[13],"of":[14,121,192],"these":[15],"disparities":[16],"within":[17],"modern":[19,79],"modeling":[21,195],"pipeline":[22],"remains":[23],"unclear.Our":[24],"study":[25],"traces":[26],"this":[27],"\"dialect":[28],"tax\"":[29],"across":[30],"natural":[32],"processing":[34],"pipeline.Using":[35],"parallel":[36],"English":[37,57],"dialect":[38,112,148,168,175],"corpora":[39],"that":[40,51,127,173],"hold":[41],"meaning":[42],"fixed":[43],"while":[44,161],"varying":[45],"surface":[46],"form,":[47],"we":[48,65],"first":[49],"confirm":[50],"LMs":[52,80],"recognize":[53],"matched":[54],"Standard":[55],"American":[56],"(SAE)":[58],"and":[59,77,90,104,179],"texts":[61,84],"as":[62],"semantically":[63,130],"equivalent.However,":[64],"discover":[66],"further":[67],"representational":[68],"corresponding":[70],"to":[71,135,153,158],"downstream":[72],"gaps.Across":[74],"model":[75],"families":[76],"generations,":[78],"still":[81],"encode":[82],"unequally":[85],"during":[86],"tokenization,":[87],"pre-training,":[88,111],"post-training,":[89,142],"inference.Strikingly,":[91],"bypassing":[92],"traditional":[93],"subword":[94],"segmentation":[95],"via":[96],"a":[97],"character-level":[98],"counterfactual":[99],"tokenizer":[100],"removes":[101],"neither":[102],"input":[103],"output":[105],"asymmetries":[106],"nor":[107],"accuracy":[109],"gaps.During":[110],"pairs":[113,120],"induce":[114],"more":[115],"divergent":[116],"gradient":[117],"updates":[118],"than":[119,138,157],"entirely":[122],"unrelated":[123,139],"SAE":[124,140],"documents,":[125],"indicating":[126],"find":[129],"equivalent":[131],"content":[133],"harder":[134],"learn":[136],"from":[137],"documents.During":[141],"reward":[143],"show":[145],"contextual,":[146],"unstable":[147],"preferences,":[149],"assigning":[150],"higher":[151],"values":[152],"isolated":[154],"AAVE-exclusive":[155],"tokens":[156],"SAE-exclusive":[159],"tokens,":[160],"full":[162],"reasoning":[163],"contexts":[164],"receive":[165],"task-and":[166],"model-dependent":[167],"penalties.Overall,":[169],"our":[170],"findings":[171],"suggest":[172],"tax":[176],"is":[177],"encoded":[178],"accumulated":[180],"not":[181],"by":[182],"any":[183],"one":[184],"step":[185,191],"isolation,":[187],"at":[189],"every":[190],"process.":[196]},"counts_by_year":[],"updated_date":"2026-11-01T11:40:01.658920","created_date":"2026-08-28T00:00:00"}