{"id":"https://openalex.org/W7162068334","doi":"https://doi.org/10.48550/arxiv.2605.21969","title":"LLM Retrieval for Stable and Predictable Ad Recommendations","display_name":"LLM Retrieval for Stable and Predictable Ad Recommendations","publication_year":2026,"publication_date":"2026-05-21","ids":{"openalex":"https://openalex.org/W7162068334","doi":"https://doi.org/10.48550/arxiv.2605.21969"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.21969","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21969","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.2605.21969","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5047033810","display_name":"Vinodh Kumar Sunkara","orcid":"https://orcid.org/0000-0002-1764-676X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sunkara, Vinodh Kumar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136798492","display_name":"Satheeshkumar Karuppusamy","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Karuppusamy, Satheeshkumar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100652505","display_name":"Heng Xu","orcid":"https://orcid.org/0000-0003-2138-3963"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Hangjun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013498616","display_name":"Sai Deepika Regani","orcid":"https://orcid.org/0000-0002-3355-6502"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Regani, Sai Deepika","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136748766","display_name":"Kshitij Gupta","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gupta, Kshitij","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5117261651","display_name":"Gaby Nahum","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nahum, Gaby","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136750891","display_name":"Sneha Iyer","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Iyer, Sneha","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031569963","display_name":"Jean-Baptiste Fiot","orcid":"https://orcid.org/0000-0002-1049-7677"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fiot, Jean-Baptiste","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136771421","display_name":"Yinglong Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Yinglong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136808441","display_name":"Xiaowen Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Xiaowen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136734714","display_name":"Atul Jangra","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jangra, Atul","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136798291","display_name":"Yucheng Liu","orcid":"https://orcid.org/0000-0002-1917-8330"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Yucheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136754360","display_name":"Jinghao Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yan, Jinghao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136749963","display_name":"Vijay Pappu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pappu, Vijay","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136770886","display_name":"Benjamin Schulte","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Schulte, Benjamin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136802979","display_name":"Deepak Chandra","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chandra, Deepak","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/T10203","display_name":"Recommender Systems and Techniques","score":0.7592999935150146,"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/T10203","display_name":"Recommender Systems and Techniques","score":0.7592999935150146,"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/T12101","display_name":"Advanced Bandit Algorithms Research","score":0.02199999988079071,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T11704","display_name":"Mobile Crowdsensing and Crowdsourcing","score":0.014800000004470348,"subfield":{"id":"https://openalex.org/subfields/1706","display_name":"Computer Science Applications"},"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/predictability","display_name":"Predictability","score":0.9004999995231628},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.6276999711990356},{"id":"https://openalex.org/keywords/recommender-system","display_name":"Recommender system","score":0.4902999997138977},{"id":"https://openalex.org/keywords/stability","display_name":"Stability (learning theory)","score":0.3977000117301941},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.3813000023365021},{"id":"https://openalex.org/keywords/generative-grammar","display_name":"Generative grammar","score":0.33180001378059387}],"concepts":[{"id":"https://openalex.org/C197640229","wikidata":"https://www.wikidata.org/wiki/Q2534066","display_name":"Predictability","level":2,"score":0.9004999995231628},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.761900007724762},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.6276999711990356},{"id":"https://openalex.org/C557471498","wikidata":"https://www.wikidata.org/wiki/Q554950","display_name":"Recommender system","level":2,"score":0.4902999997138977},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.44040000438690186},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43869999051094055},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.3977000117301941},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.382099986076355},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.3813000023365021},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.33180001378059387},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3158000111579895},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3125},{"id":"https://openalex.org/C81669768","wikidata":"https://www.wikidata.org/wiki/Q2359161","display_name":"Precision and recall","level":2,"score":0.2890999913215637},{"id":"https://openalex.org/C115174607","wikidata":"https://www.wikidata.org/wiki/Q1100934","display_name":"Click-through rate","level":2,"score":0.2770000100135803}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.21969","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21969","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.2605.21969","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21969","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":"Industry, innovation and infrastructure","id":"https://metadata.un.org/sdg/9","score":0.4584096670150757}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Traditional":[0],"ads":[1,33,103,189,196,221],"recommendation":[2,197,236],"systems":[3,239],"have":[4],"primarily":[5],"focused":[6],"on":[7],"optimizing":[8],"for":[9,96,155],"prediction":[10,42,51],"accuracy":[11],"of":[12,32,72,101,134,165],"click":[13],"or":[14,23],"conversion":[15],"events":[16],"using":[17],"canonical":[18],"metrics":[19,128],"such":[20,80],"as":[21,81,152],"recall":[22],"normalized":[24],"discounted":[25],"cumulative":[26],"gain":[27],"(NDCG).":[28],"With":[29],"the":[30,41,70,132,135,153,159,174,183,220],"hyper-growth":[31],"inventory":[34],"and":[35,44,53,85,99,106,178,204,213,237,243],"liquidity":[36],"with":[37,62],"generative":[38],"AI":[39],"technologies,":[40],"stability":[43,52,98],"predictability":[45,54,100,212,244],"is":[46,224],"becoming":[47],"increasingly":[48],"critical.":[49],"Intuitively,":[50],"can":[55,229],"be":[56,230],"defined":[57],"to":[58,64,76,146,182,233],"quantify":[59],"system":[60],"robustness":[61],"respect":[63],"minor/noisy":[65],"input":[66],"(ads,":[67],"creatives)":[68],"perturbations,":[69],"lack":[71],"which":[73,150],"could":[74],"lead":[75],"advertiser":[77,175],"perceivable":[78],"problems":[79],"repeatability,":[82],"cold":[83],"start":[84],"under-exploration.":[86],"In":[87],"this":[88,187,223],"paper,":[89],"we":[90],"introduce":[91],"a":[92,193,225],"new":[93],"evaluation":[94],"framework":[95,114,191,227],"quantifying":[97],"an":[102,108,166],"recommender":[104],"system,":[105,198],"present":[107],"online":[109,205],"validated":[110],"semantic":[111,141,163],"candidate":[112],"generation":[113],"powered":[115],"by":[116,129],"fine-tuned":[117],"Large":[118],"Language":[119],"Models":[120],"(LLMs)":[121],"that":[122,169,228],"showed":[123],"significant":[124,200],"improvement":[125],"along":[126],"these":[127],"fundamentally":[130],"improving":[131],"semantic-awareness":[133],"system.":[136],"The":[137],"approach":[138],"extracts":[139],"hierarchical":[140],"attributes":[142],"from":[143,173],"ad":[144],"creatives":[145],"obtain":[147],"LLM":[148,188],"representations,":[149],"serve":[151],"foundation":[154],"graph-based":[156],"expansion,":[157],"ensuring":[158],"retrieved":[160],"candidates":[161],"encapsulate":[162],"variants":[164,172],"ad,":[167],"guaranteeing":[168],"small":[170],"creative":[171],"yield":[176],"consistent":[177],"explainable":[179],"delivery":[180],"results":[181],"user.":[184],"We":[185],"tested":[186],"retrieval":[190,238],"in":[192,210,219],"large-scale":[194,235],"industrial":[195],"demonstrating":[199],"improvements":[201],"across":[202],"offline":[203],"A/B":[206],"experiments,":[207],"showcasing":[208],"gains":[209],"both":[211],"traditional":[214],"performance":[215],"metrics.":[216],"Although":[217],"evaluated":[218],"stack,":[222],"general":[226],"applied":[231],"broadly":[232],"any":[234],"facing":[240],"similar":[241],"scaling":[242],"challenges.":[245]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-23T00:00:00"}