{"id":"https://openalex.org/W7161944239","doi":"https://doi.org/10.48550/arxiv.2605.20693","title":"Interpretable Discriminative Text Representations via Agreement and Label Disentanglement","display_name":"Interpretable Discriminative Text Representations via Agreement and Label Disentanglement","publication_year":2026,"publication_date":"2026-05-20","ids":{"openalex":"https://openalex.org/W7161944239","doi":"https://doi.org/10.48550/arxiv.2605.20693"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.20693","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20693","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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.2605.20693","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136702308","display_name":"TONG WANG","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Tong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136695349","display_name":"Yiqing Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yiqing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136714400","display_name":"Leo Yang Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Leo Yang","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/T10664","display_name":"Sentiment Analysis and Opinion Mining","score":0.21170000731945038,"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/T10664","display_name":"Sentiment Analysis and Opinion Mining","score":0.21170000731945038,"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/T11550","display_name":"Text and Document Classification Technologies","score":0.1200999990105629,"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":0.10599999874830246,"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/discriminative-model","display_name":"Discriminative model","score":0.8378999829292297},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.65829998254776},{"id":"https://openalex.org/keywords/agreement","display_name":"Agreement","score":0.48539999127388},{"id":"https://openalex.org/keywords/paraphrase","display_name":"Paraphrase","score":0.4487000107765198},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.43479999899864197},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.4122999906539917},{"id":"https://openalex.org/keywords/reliability","display_name":"Reliability (semiconductor)","score":0.4115000069141388},{"id":"https://openalex.org/keywords/meaning","display_name":"Meaning (existential)","score":0.3921000063419342}],"concepts":[{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.8378999829292297},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.711899995803833},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6694999933242798},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.65829998254776},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5699999928474426},{"id":"https://openalex.org/C2776818064","wikidata":"https://www.wikidata.org/wiki/Q829903","display_name":"Agreement","level":2,"score":0.48539999127388},{"id":"https://openalex.org/C2780922921","wikidata":"https://www.wikidata.org/wiki/Q255189","display_name":"Paraphrase","level":2,"score":0.4487000107765198},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.43479999899864197},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.4122999906539917},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.4115000069141388},{"id":"https://openalex.org/C2780876879","wikidata":"https://www.wikidata.org/wiki/Q3054749","display_name":"Meaning (existential)","level":2,"score":0.3921000063419342},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.35190001130104065},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.35179999470710754},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.34769999980926514},{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.32420000433921814},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.3082999885082245},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.29490000009536743},{"id":"https://openalex.org/C31170391","wikidata":"https://www.wikidata.org/wiki/Q188619","display_name":"Hierarchy","level":2,"score":0.29159998893737793},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.2840000092983246},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.259799987077713},{"id":"https://openalex.org/C52622490","wikidata":"https://www.wikidata.org/wiki/Q1026626","display_name":"Feature extraction","level":2,"score":0.2563999891281128},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.25440001487731934},{"id":"https://openalex.org/C100660578","wikidata":"https://www.wikidata.org/wiki/Q18733","display_name":"Recall","level":2,"score":0.2508000135421753}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.20693","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20693","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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.2605.20693","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20693","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":[{"display_name":"Reduced inequalities","score":0.7566075921058655,"id":"https://metadata.un.org/sdg/10"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Interpretable":[0],"text":[1,59,112,161,212],"representations":[2,22],"should":[3,63,84],"expose":[4],"coordinates":[5,204],"that":[6,40,103,178,201],"are":[7,43],"not":[8,85],"only":[9],"predictive,":[10],"but":[11],"also":[12],"meaningful":[13],"enough":[14],"for":[15,56,210],"independent":[16,72],"auditors":[17],"to":[18,36,135],"apply.":[19],"Existing":[20],"discriminative":[21,58],"often":[23],"use":[24],"anonymous":[25],"embedding":[26],"directions,":[27],"while":[28,164],"concept-bottleneck":[29],"and":[30,78,106,120,168,184,190],"LLM-assisted":[31,96],"methods":[32],"attach":[33],"natural-language":[34],"names":[35],"features":[37,108,122,180],"without":[38],"ensuring":[39],"those":[41],"definitions":[42],"reproducible":[44],"or":[45],"distinct":[46],"from":[47,109],"the":[48,75,82,88,132,155],"target":[49],"label.":[50],"We":[51,91],"propose":[52],"an":[53,100],"operational":[54],"criterion":[55,94],"interpretable":[57,211],"representations:":[60],"each":[61],"coordinate":[62],"satisfy":[64],"conceptual":[65],"clarity,":[66],"measured":[67],"by":[68,123],"chance-adjusted":[69],"agreement":[70,141,186],"between":[71],"annotators":[73],"applying":[74],"feature":[76,83],"definition,":[77],"label":[79],"disentanglement,":[80],"meaning":[81],"merely":[86],"paraphrase":[87],"prediction":[89],"target.":[90],"instantiate":[92],"this":[93],"in":[95],"Feature":[97],"Discovery":[98],"(LFD),":[99],"iterative":[101],"method":[102],"proposes":[104],"lexical":[105],"semantic":[107],"contrastive":[110],"outcome-opposed":[111],"pairs,":[113],"screens":[114],"candidates":[115],"using":[116],"cross-LLM":[117],"Cohen's":[118],"$\u03ba$,":[119],"selects":[121],"residual":[124],"held-out":[125],"predictive":[126,156],"gain.":[127],"A":[128],"stylized":[129],"analysis":[130],"connects":[131],"$\u03ba$":[133],"screen":[134],"a":[136,143,159,206],"per-feature":[137],"annotation-noise":[138],"bound,":[139],"formalizing":[140],"as":[142,195],"reliability":[144],"check.":[145],"Across":[146],"ten":[147],"text-classification":[148],"tasks":[149],"spanning":[150],"seven":[151],"corpora,":[152],"LFD":[153,179],"matches":[154],"performance":[157],"of":[158],"strong":[160],"bottleneck":[162],"baseline":[163,188],"producing":[165],"substantially":[166],"clearer":[167],"less":[169,196],"label-entangled":[170],"features.":[171],"Human":[172],"audits":[173],"with":[174],"232":[175],"raters":[176,191],"show":[177],"achieve":[181],"higher":[182],"human--human":[183],"human--LLM":[185],"than":[187],"concepts,":[189],"consistently":[192],"judge":[193],"them":[194],"label-leaking.":[197],"These":[198],"results":[199],"suggest":[200],"agreement-tested,":[202],"label-disentangled":[203],"provide":[205],"practical":[207],"auditability":[208],"standard":[209],"classification.":[213]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-22T00:00:00"}