{"id":"https://openalex.org/W4415367566","doi":"https://doi.org/10.1109/isit63088.2025.11195330","title":"Leveraging Conditional Mutual Information to Improve Large Language Model Fine-Tuning for Classification","display_name":"Leveraging Conditional Mutual Information to Improve Large Language Model Fine-Tuning for Classification","publication_year":2025,"publication_date":"2025-06-22","ids":{"openalex":"https://openalex.org/W4415367566","doi":"https://doi.org/10.1109/isit63088.2025.11195330"},"language":null,"primary_location":{"id":"doi:10.1109/isit63088.2025.11195330","is_oa":false,"landing_page_url":"https://doi.org/10.1109/isit63088.2025.11195330","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Symposium on Information Theory (ISIT)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5114426758","display_name":"Thanushon Sivakaran","orcid":null},"institutions":[{"id":"https://openalex.org/I151746483","display_name":"University of Waterloo","ror":"https://ror.org/01aff2v68","country_code":"CA","type":"education","lineage":["https://openalex.org/I151746483"]}],"countries":["CA"],"is_corresponding":true,"raw_author_name":"Thanushon Sivakaran","raw_affiliation_strings":["University of Waterloo,Department of Electrical and Computer Engineering,Waterloo,Ontario"],"affiliations":[{"raw_affiliation_string":"University of Waterloo,Department of Electrical and Computer Engineering,Waterloo,Ontario","institution_ids":["https://openalex.org/I151746483"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5080794633","display_name":"En\u2010hui Yang","orcid":"https://orcid.org/0000-0003-1504-7584"},"institutions":[{"id":"https://openalex.org/I151746483","display_name":"University of Waterloo","ror":"https://ror.org/01aff2v68","country_code":"CA","type":"education","lineage":["https://openalex.org/I151746483"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"En-Hui Yang","raw_affiliation_strings":["University of Waterloo,Department of Electrical and Computer Engineering,Waterloo,Ontario"],"affiliations":[{"raw_affiliation_string":"University of Waterloo,Department of Electrical and Computer Engineering,Waterloo,Ontario","institution_ids":["https://openalex.org/I151746483"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":["https://openalex.org/A5114426758"],"corresponding_institution_ids":["https://openalex.org/I151746483"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.16439007,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"6"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.967199981212616,"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/T10028","display_name":"Topic Modeling","score":0.967199981212616,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9578999876976013,"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/mutual-information","display_name":"Mutual information","score":0.72079998254776},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.5656999945640564},{"id":"https://openalex.org/keywords/adaptability","display_name":"Adaptability","score":0.5397999882698059},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4523000121116638},{"id":"https://openalex.org/keywords/information-theory","display_name":"Information theory","score":0.34619998931884766},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.33079999685287476}],"concepts":[{"id":"https://openalex.org/C152139883","wikidata":"https://www.wikidata.org/wiki/Q252973","display_name":"Mutual information","level":2,"score":0.72079998254776},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7014999985694885},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6074000000953674},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5656999945640564},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5490999817848206},{"id":"https://openalex.org/C177606310","wikidata":"https://www.wikidata.org/wiki/Q5674297","display_name":"Adaptability","level":2,"score":0.5397999882698059},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4523000121116638},{"id":"https://openalex.org/C52622258","wikidata":"https://www.wikidata.org/wiki/Q131222","display_name":"Information theory","level":2,"score":0.34619998931884766},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.33079999685287476},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.32100000977516174},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.28619998693466187},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2676999866962433},{"id":"https://openalex.org/C75294576","wikidata":"https://www.wikidata.org/wiki/Q5165192","display_name":"Contextual image classification","level":3,"score":0.2597000002861023},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.25119999051094055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/isit63088.2025.11195330","is_oa":false,"landing_page_url":"https://doi.org/10.1109/isit63088.2025.11195330","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Symposium on Information Theory (ISIT)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":8,"referenced_works":["https://openalex.org/W1995875735","https://openalex.org/W2130433404","https://openalex.org/W2130956967","https://openalex.org/W2294370754","https://openalex.org/W3001434439","https://openalex.org/W3034999214","https://openalex.org/W4245857357","https://openalex.org/W4401692272"],"related_works":[],"abstract_inverted_index":{"Although":[0],"large":[1],"language":[2,179],"models":[3],"(LLMs)":[4],"have":[5],"demonstrated":[6],"remarkable":[7],"capabilities":[8],"in":[9,45,73,124,128],"recent":[10],"years,":[11],"the":[12,27,78,120,165],"potential":[13,153],"of":[14,31,108,130],"information":[15,28,168],"theory":[16,169],"(IT)":[17],"to":[18,36,51,60,114,136],"enhance":[19,61],"LLM":[20,37,74,99,160,171],"development":[21],"remains":[22],"underexplored.":[23],"This":[24],"paper":[25],"introduces":[26],"theoretic":[29],"principle":[30],"Conditional":[32],"Mutual":[33],"Information":[34],"(CMI)":[35],"fine-tuning":[38],"for":[39,65,89,143,158,176],"classification":[40,111,133],"tasks,":[41],"exploring":[42],"its":[43,152],"promise":[44],"two":[46],"main":[47],"ways:":[48],"minimizing":[49,96],"CMI":[50,59,72,97,118],"improve":[52],"a":[53,155],"model's":[54],"standalone":[55,146],"performance":[56,104,126],"and":[57,148,170],"maximizing":[58,117],"knowledge":[62],"distillation":[63],"(KD)":[64],"more":[66],"capable":[67],"student":[68,149],"models.":[69,180],"To":[70],"apply":[71],"fine-tuning,":[75,100],"we":[76,101],"adapt":[77],"recently":[79],"proposed":[80],"CMI-constrained":[81],"deep":[82],"learning":[83],"framework,":[84],"which":[85],"was":[86],"initially":[87],"developed":[88],"image":[90],"classification,":[91],"with":[92],"some":[93],"modification.":[94],"By":[95],"during":[98,119],"achieve":[102],"superior":[103],"gains":[105],"on":[106],"6":[107,129],"8":[109,131],"GLUE":[110,132],"tasks":[112,134],"compared":[113,135],"BERT.":[115],"Additionally,":[116],"KD":[121],"process":[122],"results":[123],"significant":[125],"improvements":[127],"DistilBERT.":[137],"These":[138],"findings":[139],"demonstrate":[140],"CMI's":[141],"adaptability":[142],"optimizing":[144],"both":[145],"LLMs":[147],"models,":[150],"showcasing":[151],"as":[154],"robust":[156],"framework":[157],"advancing":[159],"fine-tuning.":[161],"Our":[162],"work":[163],"bridges":[164],"gap":[166],"between":[167],"development,":[172],"offering":[173],"new":[174],"insights":[175],"building":[177],"high-performing":[178]},"counts_by_year":[],"updated_date":"2026-03-07T16:01:11.037858","created_date":"2025-10-21T00:00:00"}