{"id":"https://openalex.org/W7156130353","doi":"https://doi.org/10.48550/arxiv.2604.22722","title":"Aligning Dense Retrievers with LLM Utility via Distillation","display_name":"Aligning Dense Retrievers with LLM Utility via Distillation","publication_year":2026,"publication_date":"2026-04-24","ids":{"openalex":"https://openalex.org/W7156130353","doi":"https://doi.org/10.48550/arxiv.2604.22722"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.22722","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.22722","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":"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.22722","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5044203866","display_name":"Rajinder Sandhu","orcid":"https://orcid.org/0000-0001-8658-6089"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sandhu, Rajinder","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134733966","display_name":"Di Mu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mu, Di","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134709013","display_name":"Cheng Chang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chang, Cheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134692013","display_name":"Md Shahriar Tasjid","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tasjid, Md Shahriar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134686051","display_name":"Himanshu Rai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rai, Himanshu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027341076","display_name":"Maksims Volkovs","orcid":"https://orcid.org/0009-0007-0561-2187"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Volkovs, Maksims","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5119087594","display_name":"\u52a1 \u560e","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Ga","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/T10286","display_name":"Information Retrieval and Search Behavior","score":0.7595000267028809,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10286","display_name":"Information Retrieval and Search Behavior","score":0.7595000267028809,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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":0.04740000143647194,"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":0.03449999913573265,"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/perplexity","display_name":"Perplexity","score":0.5997999906539917},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.5544000267982483},{"id":"https://openalex.org/keywords/merge","display_name":"Merge (version control)","score":0.4465999901294708},{"id":"https://openalex.org/keywords/discriminator","display_name":"Discriminator","score":0.34630000591278076},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.3441999852657318},{"id":"https://openalex.org/keywords/vector-space-model","display_name":"Vector space model","score":0.3330000042915344},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.31529998779296875}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7336999773979187},{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.5997999906539917},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.5544000267982483},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.538100004196167},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.4465999901294708},{"id":"https://openalex.org/C2779803651","wikidata":"https://www.wikidata.org/wiki/Q5282088","display_name":"Discriminator","level":3,"score":0.34630000591278076},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.3441999852657318},{"id":"https://openalex.org/C89686163","wikidata":"https://www.wikidata.org/wiki/Q1187982","display_name":"Vector space model","level":2,"score":0.3330000042915344},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.32659998536109924},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.31529998779296875},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.30820000171661377},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.2996000051498413},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.2994999885559082},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.29580000042915344},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.28029999136924744},{"id":"https://openalex.org/C1667742","wikidata":"https://www.wikidata.org/wiki/Q10927554","display_name":"Image retrieval","level":3,"score":0.26840001344680786},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.257999986410141}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.22722","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.22722","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.22722","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.22722","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":"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Dense":[0],"vector":[1],"retrieval":[2,58,62,107,143],"is":[3,127],"the":[4,93,102,120,132],"practical":[5],"backbone":[6],"of":[7],"Retrieval-":[8],"Augmented":[9],"Generation":[10],"(RAG),":[11],"but":[12,30],"similarity":[13],"search":[14],"can":[15],"suffer":[16],"from":[17,77],"precision":[18],"limitations.":[19],"Conversely,":[20],"utility-based":[21],"approaches":[22],"leveraging":[23],"LLM":[24,99,134],"re-ranking":[25,135],"often":[26],"achieve":[27],"superior":[28],"performance":[29],"are":[31],"computationally":[32],"prohibitive":[33],"and":[34,114],"prone":[35],"to":[36,50,71],"noise":[37],"inherent":[38],"in":[39],"perplexity":[40,78],"estimation.":[41],"We":[42,60],"propose":[43],"Utility-Aligned":[44],"Embeddings":[45],"(UAE),":[46],"a":[47,55,64,69,73,81],"framework":[48],"designed":[49],"merge":[51],"these":[52],"advantages":[53],"into":[54,92],"practical,":[56],"high-performance":[57],"method.":[59],"formulate":[61],"as":[63],"distribution":[65,75],"matching":[66],"problem,":[67],"training":[68],"bi-encoder":[70],"imitate":[72],"utility":[74,89,146],"derived":[76],"reduction":[79],"using":[80],"Utility-Modulated":[82],"InfoNCE":[83],"objective.":[84],"This":[85],"approach":[86],"injects":[87],"graded":[88],"signals":[90],"directly":[91],"embedding":[94],"space":[95],"without":[96],"requiring":[97],"test-time":[98],"inference.":[100],"On":[101],"QASPER":[103],"benchmark,":[104],"UAE":[105,126],"improves":[106],"Recall@1":[108],"by":[109,112,117],"30.59%,":[110],"MAP":[111],"30.16%":[113],"Token":[115],"F1":[116],"17.3%":[118],"over":[119,128],"strong":[121],"semantic":[122],"baseline":[123],"BGE-Base.":[124],"Crucially,":[125],"180x":[129],"faster":[130],"than":[131],"efficient":[133],"methods":[136],"preserving":[137],"competitive":[138],"performance,":[139],"demonstrating":[140],"that":[141],"aligning":[142],"with":[144],"generative":[145],"yields":[147],"reliable":[148],"contexts":[149],"at":[150],"scale.":[151]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-28T00:00:00"}