{"id":"https://openalex.org/W7168240740","doi":"https://doi.org/10.48550/arxiv.2607.10096","title":"Scaling and Stabilizing Large-Scale Embedding-Based Retrieval","display_name":"Scaling and Stabilizing Large-Scale Embedding-Based Retrieval","publication_year":2026,"publication_date":"2026-07-11","ids":{"openalex":"https://openalex.org/W7168240740","doi":"https://doi.org/10.48550/arxiv.2607.10096"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2607.10096","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.10096","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,"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.2607.10096","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5140717152","display_name":"Zhen Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Zhen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140593325","display_name":"Juexin Lin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Juexin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140717736","display_name":"Hongwei Shang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shang, Hongwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140749156","display_name":"Kaihao Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Kaihao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140623599","display_name":"Feng Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Feng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140748745","display_name":"Satya Chembolu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chembolu, Satya","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140642675","display_name":"Xunfan Cai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cai, Xunfan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140693988","display_name":"Xinyi Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Xinyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140628496","display_name":"Cun Mu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mu, Cun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140617490","display_name":"Tony Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Tony","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5140649188","display_name":"Ciya Liao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liao, Ciya","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":0.5715000033378601,"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":0.5715000033378601,"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/T10203","display_name":"Recommender Systems and Techniques","score":0.07450000196695328,"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/T11550","display_name":"Text and Document Classification Technologies","score":0.04390000179409981,"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/pipeline","display_name":"Pipeline (software)","score":0.6802999973297119},{"id":"https://openalex.org/keywords/heuristics","display_name":"Heuristics","score":0.6021000146865845},{"id":"https://openalex.org/keywords/metadata","display_name":"Metadata","score":0.5541999936103821},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.4674000144004822},{"id":"https://openalex.org/keywords/lift","display_name":"Lift (data mining)","score":0.4415999948978424},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.43950000405311584},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.37229999899864197},{"id":"https://openalex.org/keywords/limit","display_name":"Limit (mathematics)","score":0.3644999861717224}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7315999865531921},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.6802999973297119},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.6021000146865845},{"id":"https://openalex.org/C93518851","wikidata":"https://www.wikidata.org/wiki/Q180160","display_name":"Metadata","level":2,"score":0.5541999936103821},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.4674000144004822},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.4578000009059906},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4424999952316284},{"id":"https://openalex.org/C139002025","wikidata":"https://www.wikidata.org/wiki/Q3001212","display_name":"Lift (data mining)","level":2,"score":0.4415999948978424},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.43950000405311584},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4185999929904938},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.37229999899864197},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.3644999861717224},{"id":"https://openalex.org/C100368936","wikidata":"https://www.wikidata.org/wiki/Q1411725","display_name":"Crawling","level":2,"score":0.3619000017642975},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.33640000224113464},{"id":"https://openalex.org/C2777230681","wikidata":"https://www.wikidata.org/wiki/Q7923820","display_name":"Vetting","level":2,"score":0.3109999895095825},{"id":"https://openalex.org/C81669768","wikidata":"https://www.wikidata.org/wiki/Q2359161","display_name":"Precision and recall","level":2,"score":0.30480000376701355},{"id":"https://openalex.org/C24756922","wikidata":"https://www.wikidata.org/wiki/Q1757694","display_name":"Data quality","level":3,"score":0.30090001225471497},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.2953000068664551},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.2766000032424927},{"id":"https://openalex.org/C551230270","wikidata":"https://www.wikidata.org/wiki/Q4368942","display_name":"Data retrieval","level":2,"score":0.27149999141693115},{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.2685000002384186},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.25999999046325684}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2607.10096","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.10096","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,"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.2607.10096","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.10096","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,"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":0.6296731233596802,"id":"https://metadata.un.org/sdg/10"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Embedding-based":[0],"retrieval":[1,72],"(EBR)":[2],"is":[3,12,191],"foundational":[4],"to":[5,56,70,118,138,149,175],"large-scale":[6],"e-commerce":[7],"search,":[8],"yet":[9],"its":[10],"effectiveness":[11],"often":[13],"constrained":[14],"by":[15,122],"the":[16,22,26,78,172,176,188],"quality":[17,100],"of":[18,25,47,49,77,125],"training":[19],"signals":[20],"and":[21,74,101,127,158,184,202],"representational":[23],"capacity":[24],"encoder.":[27,153],"Standard":[28],"dual-encoders":[29],"suffer":[30],"from":[31,147,171],"a":[32,65,75,90,150,156,163,197,203],"training-inference":[33],"gap:":[34],"they":[35],"are":[36,106],"optimized":[37],"on":[38],"narrow":[39],"candidate":[40],"pools":[41],"but":[42],"must":[43],"discriminate":[44],"against":[45],"hundreds":[46],"millions":[48],"items":[50],"during":[51],"inference.":[52],"Furthermore,":[53],"while":[54],"transitioning":[55],"higher-capacity":[57,151],"backbones":[58],"can":[59,68],"mitigate":[60],"this":[61,86],"gap,":[62],"simply":[63],"replacing":[64],"mature":[66],"model":[67,102,174],"lead":[69],"inconsistent":[71],"behavior":[73],"loss":[76],"domain-specific":[79,169],"knowledge":[80],"established":[81],"in":[82,193,200,206],"previous":[83],"iterations.":[84],"In":[85],"paper,":[87],"we":[88,161],"present":[89],"unified":[91],"pipeline":[92,190],"deployed":[93,192],"at":[94],"Walmart":[95],"that":[96,167],"addresses":[97],"both":[98],"signal":[99],"evolution.":[103],"Our":[104],"contributions":[105],"two-fold:":[107],"(1)":[108],"Hybrid":[109,128],"Hard":[110],"Negative":[111],"Mining:":[112],"We":[113,145],"integrate":[114],"Online":[115],"Cross-Batch":[116],"Sampling":[117],"increase":[119],"negative":[120],"diversity":[121],"an":[123],"order":[124],"magnitude":[126],"Offline":[129],"Mining,":[130],"which":[131],"combines":[132],"cross-encoder":[133],"predictions":[134],"with":[135],"metadata":[136],"heuristics":[137],"identify":[139],"nuanced":[140],"mismatches.":[141],"(2)":[142],"Legacy-Aware":[143],"Distillation:":[144],"transition":[146],"DistilBERT":[148],"GTE-base":[152],"To":[154],"ensure":[155],"smooth":[157],"superior":[159],"transition,":[160],"introduce":[162],"Warm-Start":[164],"Distillation":[165],"technique":[166],"transfers":[168],"expertise":[170],"legacy":[173],"new":[177],"backbone.":[178],"Validated":[179],"through":[180],"extensive":[181],"offline":[182],"experiments":[183],"online":[185],"A/B":[186],"testing,":[187],"proposed":[189],"live":[194],"production,":[195],"delivering":[196],"+7.34%":[198],"improvement":[199],"NDCG@5":[201],"+0.50%":[204],"lift":[205],"gross":[207],"revenue.":[208]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-07-15T00:00:00"}