{"id":"https://openalex.org/W4414034849","doi":"https://doi.org/10.1145/3705328.3748033","title":"Correcting the LogQ Correction: Revisiting Sampled Softmax for Large-Scale Retrieval","display_name":"Correcting the LogQ Correction: Revisiting Sampled Softmax for Large-Scale Retrieval","publication_year":2025,"publication_date":"2025-09-06","ids":{"openalex":"https://openalex.org/W4414034849","doi":"https://doi.org/10.1145/3705328.3748033"},"language":"en","primary_location":{"id":"doi:10.1145/3705328.3748033","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3705328.3748033","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Nineteenth ACM Conference on Recommender Systems","raw_type":"proceedings-article"},"type":"article","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2507.09331","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5091814356","display_name":"Kirill Khrylchenko","orcid":"https://orcid.org/0009-0007-3640-8795"},"institutions":[{"id":"https://openalex.org/I58957048","display_name":"Yandex (Russia)","ror":"https://ror.org/04dbch786","country_code":"RU","type":"company","lineage":["https://openalex.org/I58957048"]}],"countries":["RU"],"is_corresponding":true,"raw_author_name":"Kirill Khrylchenko","raw_affiliation_strings":["Yandex, Moscow, Russian Federation"],"raw_orcid":"https://orcid.org/0009-0007-3640-8795","affiliations":[{"raw_affiliation_string":"Yandex, Moscow, Russian Federation","institution_ids":["https://openalex.org/I58957048"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031658527","display_name":"Vladimir Baikalov","orcid":"https://orcid.org/0009-0009-4864-2305"},"institutions":[{"id":"https://openalex.org/I58957048","display_name":"Yandex (Russia)","ror":"https://ror.org/04dbch786","country_code":"RU","type":"company","lineage":["https://openalex.org/I58957048"]}],"countries":["RU"],"is_corresponding":false,"raw_author_name":"Vladimir Baikalov","raw_affiliation_strings":["Yandex, Moscow, Russian Federation"],"raw_orcid":"https://orcid.org/0009-0009-4864-2305","affiliations":[{"raw_affiliation_string":"Yandex, Moscow, Russian Federation","institution_ids":["https://openalex.org/I58957048"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5068390397","display_name":"Sergei Makeev","orcid":null},"institutions":[{"id":"https://openalex.org/I58957048","display_name":"Yandex (Russia)","ror":"https://ror.org/04dbch786","country_code":"RU","type":"company","lineage":["https://openalex.org/I58957048"]}],"countries":["RU"],"is_corresponding":false,"raw_author_name":"Sergei Makeev","raw_affiliation_strings":["Yandex, Moscow, Russian Federation"],"raw_orcid":"https://orcid.org/0009-0003-5451-6475","affiliations":[{"raw_affiliation_string":"Yandex, Moscow, Russian Federation","institution_ids":["https://openalex.org/I58957048"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022052407","display_name":"Artem Matveev","orcid":null},"institutions":[{"id":"https://openalex.org/I58957048","display_name":"Yandex (Russia)","ror":"https://ror.org/04dbch786","country_code":"RU","type":"company","lineage":["https://openalex.org/I58957048"]}],"countries":["RU"],"is_corresponding":false,"raw_author_name":"Artem Matveev","raw_affiliation_strings":["Yandex, Moscow, Russian Federation"],"raw_orcid":"https://orcid.org/0009-0004-0271-221X","affiliations":[{"raw_affiliation_string":"Yandex, Moscow, Russian Federation","institution_ids":["https://openalex.org/I58957048"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5119556708","display_name":"Sergei Liamaev","orcid":null},"institutions":[{"id":"https://openalex.org/I58957048","display_name":"Yandex (Russia)","ror":"https://ror.org/04dbch786","country_code":"RU","type":"company","lineage":["https://openalex.org/I58957048"]}],"countries":["RU"],"is_corresponding":false,"raw_author_name":"Sergei Liamaev","raw_affiliation_strings":["Yandex, Moscow, Russian Federation"],"raw_orcid":"https://orcid.org/0009-0009-6316-1091","affiliations":[{"raw_affiliation_string":"Yandex, Moscow, Russian Federation","institution_ids":["https://openalex.org/I58957048"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5091814356"],"corresponding_institution_ids":["https://openalex.org/I58957048"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.34200231,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"545","last_page":"550"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10203","display_name":"Recommender Systems and Techniques","score":0.9991000294685364,"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":0.9991000294685364,"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":0.998199999332428,"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/T10824","display_name":"Image Retrieval and Classification Techniques","score":0.9973000288009644,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/softmax-function","display_name":"Softmax function","score":0.7162693738937378},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6736178398132324},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.6052069664001465},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4239116609096527},{"id":"https://openalex.org/keywords/information-retrieval","display_name":"Information retrieval","score":0.3924521207809448},{"id":"https://openalex.org/keywords/geography","display_name":"Geography","score":0.08506631851196289},{"id":"https://openalex.org/keywords/cartography","display_name":"Cartography","score":0.06934446096420288},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.06602096557617188}],"concepts":[{"id":"https://openalex.org/C188441871","wikidata":"https://www.wikidata.org/wiki/Q7554146","display_name":"Softmax function","level":3,"score":0.7162693738937378},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6736178398132324},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.6052069664001465},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4239116609096527},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.3924521207809448},{"id":"https://openalex.org/C205649164","wikidata":"https://www.wikidata.org/wiki/Q1071","display_name":"Geography","level":0,"score":0.08506631851196289},{"id":"https://openalex.org/C58640448","wikidata":"https://www.wikidata.org/wiki/Q42515","display_name":"Cartography","level":1,"score":0.06934446096420288},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.06602096557617188}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1145/3705328.3748033","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3705328.3748033","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Nineteenth ACM Conference on Recommender Systems","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2507.09331","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2507.09331","pdf_url":"https://arxiv.org/pdf/2507.09331","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2507.09331","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2507.09331","pdf_url":"https://arxiv.org/pdf/2507.09331","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":23,"referenced_works":["https://openalex.org/W2080234606","https://openalex.org/W2110953678","https://openalex.org/W2136189984","https://openalex.org/W2219888463","https://openalex.org/W2512971201","https://openalex.org/W2741544350","https://openalex.org/W2783272285","https://openalex.org/W2963341956","https://openalex.org/W2963367478","https://openalex.org/W2964369530","https://openalex.org/W2972801466","https://openalex.org/W2984100107","https://openalex.org/W3023045848","https://openalex.org/W3088694469","https://openalex.org/W3154199194","https://openalex.org/W3166125679","https://openalex.org/W4290927925","https://openalex.org/W4291127187","https://openalex.org/W4384648390","https://openalex.org/W4385849288","https://openalex.org/W4386729798","https://openalex.org/W4389675217","https://openalex.org/W4401857632"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W3107204728","https://openalex.org/W4287591324","https://openalex.org/W3108503355","https://openalex.org/W3090555870","https://openalex.org/W4226420367","https://openalex.org/W2962876041","https://openalex.org/W3022820045"],"abstract_inverted_index":{"Two-tower":[0],"neural":[1],"networks":[2],"are":[3,16,73,97],"a":[4,20,54,84,105,151,182,207],"popular":[5,95,106],"architecture":[6],"for":[7,213],"the":[8,24,31,50,78,92,114,120,127,135,138,147,165,172,187,191,225,229,234,251],"retrieval":[9],"stage":[10],"in":[11,28,77,91,126,137,190],"recommender":[12],"systems.":[13],"These":[14],"models":[15],"typically":[17],"trained":[18],"with":[19,202],"softmax":[21,40,52],"loss":[22,217],"over":[23,250],"item":[25,32,124,189],"catalog.":[26],"However,":[27,81],"web-scale":[29],"settings,":[30],"catalog":[33],"is":[34,45,66,131,193,199],"often":[35],"prohibitively":[36],"large,":[37],"making":[38],"full":[39,51],"infeasible.":[41],"A":[42],"common":[43],"solution":[44],"sampled":[46,58],"softmax,":[47],"which":[48],"approximates":[49],"using":[53,146],"small":[55],"number":[56],"of":[57,122,174,231],"negatives.":[59],"One":[60],"practical":[61],"and":[62,140,177,244],"widely":[63],"adopted":[64],"approach":[65,156],"to":[67],"use":[68],"in-batch":[69,148],"negatives,":[70],"where":[71],"negatives":[72],"drawn":[74],"from":[75],"items":[76,86],"current":[79,235],"mini-batch.":[80],"this":[82,103,155,168],"introduces":[83,218],"bias:":[85],"that":[87,179,211,223],"appear":[88],"more":[89,99],"frequently":[90],"batch":[93],"(i.e.,":[94],"items)":[96],"penalized":[98],"heavily.":[100],"To":[101],"mitigate":[102],"issue,":[104],"industry":[107],"technique":[108],"known":[109],"as":[110,150],"logQ":[111,175,253],"correction":[112,130,176,209],"adjusts":[113],"logits":[115],"during":[116],"training":[117],"by":[118,133],"subtracting":[119],"log-probability":[121],"an":[123,219],"appearing":[125],"batch.":[128],"This":[129],"derived":[132],"analyzing":[134],"bias":[136],"gradient":[139],"applying":[141],"importance":[142],"sampling,":[143],"effectively":[144],"twice,":[145],"distribution":[149],"proposal":[152],"distribution.":[153],"While":[154],"improves":[157],"model":[158],"quality,":[159],"it":[160,180,198],"does":[161],"not":[162,194],"fully":[163],"eliminate":[164],"bias.":[166],"In":[167],"work,":[169],"we":[170],"revisit":[171],"derivation":[173],"show":[178],"overlooks":[181],"subtle":[183],"but":[184],"important":[185],"detail:":[186],"positive":[188],"denominator":[192],"Monte":[195],"Carlo-sampled":[196],"-":[197,228],"always":[200],"present":[201],"probability":[203,230],"1.":[204],"We":[205,237],"propose":[206],"refined":[208],"formula":[210],"accounts":[212],"this.":[214],"Notably,":[215],"our":[216,239],"interpretable":[220],"sample":[221],"weight":[222],"reflects":[224],"model's":[226],"uncertainty":[227],"misclassification":[232],"under":[233],"parameters.":[236],"evaluate":[238],"method":[240],"on":[241],"both":[242],"public":[243],"proprietary":[245],"datasets,":[246],"demonstrating":[247],"consistent":[248],"improvements":[249],"standard":[252],"correction.":[254]},"counts_by_year":[],"updated_date":"2026-03-27T05:58:40.876381","created_date":"2025-10-10T00:00:00"}