{"id":"https://openalex.org/W4412876851","doi":"https://doi.org/10.1145/3711896.3737120","title":"Self-Regularization with Sparse Autoencoders for Controllable LLM-based Classification","display_name":"Self-Regularization with Sparse Autoencoders for Controllable LLM-based Classification","publication_year":2025,"publication_date":"2025-08-03","ids":{"openalex":"https://openalex.org/W4412876851","doi":"https://doi.org/10.1145/3711896.3737120"},"language":"en","primary_location":{"id":"doi:10.1145/3711896.3737120","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3711896.3737120","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3711896.3737120","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 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3711896.3737120","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5089884284","display_name":"Xuansheng Wu","orcid":"https://orcid.org/0000-0002-7816-7658"},"institutions":[{"id":"https://openalex.org/I165733156","display_name":"University of Georgia","ror":"https://ror.org/00te3t702","country_code":"US","type":"education","lineage":["https://openalex.org/I165733156"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Xuansheng Wu","raw_affiliation_strings":["University of Georgia, Athens, Georgia, USA"],"raw_orcid":"https://orcid.org/0000-0002-7816-7658","affiliations":[{"raw_affiliation_string":"University of Georgia, Athens, Georgia, USA","institution_ids":["https://openalex.org/I165733156"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5028597555","display_name":"Wenhao Yu","orcid":"https://orcid.org/0000-0002-4075-5980"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wenhao Yu","raw_affiliation_strings":["Tencent AI Lab, Seattle, WA, USA"],"raw_orcid":"https://orcid.org/0000-0002-4075-5980","affiliations":[{"raw_affiliation_string":"Tencent AI Lab, Seattle, WA, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013379229","display_name":"Xiaoming Zha\u0131","orcid":"https://orcid.org/0000-0003-4519-1931"},"institutions":[{"id":"https://openalex.org/I165733156","display_name":"University of Georgia","ror":"https://ror.org/00te3t702","country_code":"US","type":"education","lineage":["https://openalex.org/I165733156"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Xiaoming Zhai","raw_affiliation_strings":["University of Georgia, Athens, GA, USA"],"raw_orcid":"https://orcid.org/0000-0003-4519-1931","affiliations":[{"raw_affiliation_string":"University of Georgia, Athens, GA, USA","institution_ids":["https://openalex.org/I165733156"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5007489034","display_name":"Ninghao Liu","orcid":"https://orcid.org/0000-0002-9170-2424"},"institutions":[{"id":"https://openalex.org/I165733156","display_name":"University of Georgia","ror":"https://ror.org/00te3t702","country_code":"US","type":"education","lineage":["https://openalex.org/I165733156"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ninghao Liu","raw_affiliation_strings":["University of Georgia, Athens, GA, USA"],"raw_orcid":"https://orcid.org/0000-0002-9170-2424","affiliations":[{"raw_affiliation_string":"University of Georgia, Athens, GA, USA","institution_ids":["https://openalex.org/I165733156"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"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":0.14005112,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"3250","last_page":"3260"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12535","display_name":"Machine Learning and Data Classification","score":0.9919999837875366,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.9919999837875366,"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/T10320","display_name":"Neural Networks and Applications","score":0.9905999898910522,"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/T11512","display_name":"Anomaly Detection Techniques and Applications","score":0.9886000156402588,"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/computer-science","display_name":"Computer science","score":0.6358583569526672},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.5836509466171265},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.5722863078117371},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.5229654908180237}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6358583569526672},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.5836509466171265},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5722863078117371},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.5229654908180237}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3711896.3737120","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3711896.3737120","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3711896.3737120","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 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3711896.3737120","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3711896.3737120","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3711896.3737120","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 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412876851.pdf","grobid_xml":"https://content.openalex.org/works/W4412876851.grobid-xml"},"referenced_works_count":29,"referenced_works":["https://openalex.org/W1537220347","https://openalex.org/W1677182931","https://openalex.org/W2040825624","https://openalex.org/W2884561390","https://openalex.org/W2903928064","https://openalex.org/W2948947170","https://openalex.org/W2950104108","https://openalex.org/W2950888501","https://openalex.org/W2963639656","https://openalex.org/W2970641574","https://openalex.org/W3010694149","https://openalex.org/W4221154538","https://openalex.org/W4225684474","https://openalex.org/W4234698323","https://openalex.org/W4297645476","https://openalex.org/W4382318449","https://openalex.org/W4385568240","https://openalex.org/W4388229648","https://openalex.org/W4389518965","https://openalex.org/W4389519291","https://openalex.org/W4389520373","https://openalex.org/W4401042295","https://openalex.org/W4404781566","https://openalex.org/W4404783416","https://openalex.org/W6602430550","https://openalex.org/W6653167809","https://openalex.org/W6759928758","https://openalex.org/W6843690198","https://openalex.org/W6851092083"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2033914206","https://openalex.org/W2042327336"],"abstract_inverted_index":{"Modern":[0],"text":[1,204],"classification":[2,26,59,126,205],"methods":[3],"heavily":[4],"rely":[5],"on":[6,120,156,163,206],"contextual":[7],"embeddings":[8,19,66],"from":[9,104],"large":[10],"language":[11],"models":[12],"(LLMs).":[13],"Compared":[14],"to":[15,39,50,71,81,100,148,197,214],"human-engineered":[16],"features,":[17,43,49,115],"these":[18,153],"provide":[20],"automatic":[21],"and":[22,69,83,132,143,173,218,223],"effective":[23,133],"representations":[24],"for":[25],"model":[27],"training.":[28],"However,":[29],"they":[30],"also":[31],"introduce":[32],"a":[33,78,96,130],"challenge:":[34],"we":[35,76,93,116,128],"lose":[36],"the":[37,56,88,110,125,137,140,144,150,160,179,185,198],"ability":[38],"manually":[40],"remove":[41,149],"unintended":[42,85,146,154],"such":[44],"as":[45],"sensitive":[46],"or":[47,54],"task-irrelevant":[48],"guarantee":[51],"regulatory":[52],"compliance":[53],"improve":[55,184],"generalizability":[57,187],"of":[58,152],"models.":[60],"This":[61,200],"limitation":[62],"arises":[63],"because":[64],"LLM":[65,89,105,207],"are":[67,193,225],"opaque":[68],"difficult":[70],"interpret.":[72],"In":[73,123],"this":[74],"paper,":[75],"propose":[77,129],"novel":[79],"framework":[80,162,182],"identify":[82],"regularize":[84],"features":[86,103,155,191,213],"in":[87],"latent":[90,106,208],"space.":[91],"Specifically,":[92],"first":[94],"pre-train":[95],"sparse":[97],"autoencoder":[98],"(SAE)":[99],"extract":[101],"interpretable":[102],"spaces.":[107],"To":[108],"ensure":[109],"SAE":[111],"can":[112,183],"capture":[113],"task-specific":[114,121],"further":[117],"fine-tune":[118],"it":[119],"datasets.":[122],"training":[124],"model,":[127],"simple":[131],"regularizer,":[134],"by":[135,188,210],"minimizing":[136],"similarity":[138],"between":[139],"classifier":[141],"weights":[142],"identified":[145],"feature,":[147],"impact":[151],"classification.":[157],"We":[158],"evaluate":[159],"proposed":[161,180],"three":[164],"real-world":[165],"tasks,":[166],"including":[167],"toxic":[168],"chat":[169],"detection,":[170],"reward":[171],"modeling,":[172],"disease":[174],"diagnosis.":[175],"Results":[176],"show":[177],"that":[178,192],"self-regularization":[181],"classifier's":[186],"regularizing":[189],"those":[190],"not":[194],"semantically":[195],"correlated":[196],"task.":[199],"work":[201],"pioneers":[202],"controllable":[203],"spaces":[209],"leveraging":[211],"interpreted":[212],"address":[215],"generalizability,":[216],"fairness,":[217],"privacy":[219],"challenges.":[220],"The":[221],"code":[222],"data":[224],"publicly":[226],"available":[227],"at":[228],"https://github.com/JacksonWuxs/Controllable_LLM_Classifier.":[229]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}