{"id":"https://openalex.org/W7134960620","doi":"https://doi.org/10.1109/icdmw69685.2025.00179","title":"Scaling Multimodal Search and Recommendation with Small Language Models via Upside-Down Reinforcement Learning","display_name":"Scaling Multimodal Search and Recommendation with Small Language Models via Upside-Down Reinforcement Learning","publication_year":2025,"publication_date":"2025-11-12","ids":{"openalex":"https://openalex.org/W7134960620","doi":"https://doi.org/10.1109/icdmw69685.2025.00179"},"language":null,"primary_location":{"id":"doi:10.1109/icdmw69685.2025.00179","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icdmw69685.2025.00179","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Data Mining Workshops (ICDMW)","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/A5128742292","display_name":"Yu-Chen Lin","orcid":null},"institutions":[{"id":"https://openalex.org/I1306409833","display_name":"Adobe Systems (United States)","ror":"https://ror.org/059tvcg64","country_code":"US","type":"company","lineage":["https://openalex.org/I1306409833"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yu-Chen Lin","raw_affiliation_strings":["Adobe"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Adobe","institution_ids":["https://openalex.org/I1306409833"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049493201","display_name":"Sanat Sharma","orcid":"https://orcid.org/0009-0003-2479-3041"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sanat Sharma","raw_affiliation_strings":["Meta"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Meta","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065785885","display_name":"Hari Manikandan","orcid":null},"institutions":[{"id":"https://openalex.org/I1306409833","display_name":"Adobe Systems (United States)","ror":"https://ror.org/059tvcg64","country_code":"US","type":"company","lineage":["https://openalex.org/I1306409833"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hari Manikandan","raw_affiliation_strings":["Adobe"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Adobe","institution_ids":["https://openalex.org/I1306409833"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101414463","display_name":"Jayant Kumar","orcid":"https://orcid.org/0009-0004-3858-5961"},"institutions":[{"id":"https://openalex.org/I1306409833","display_name":"Adobe Systems (United States)","ror":"https://ror.org/059tvcg64","country_code":"US","type":"company","lineage":["https://openalex.org/I1306409833"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jayant Kumar","raw_affiliation_strings":["Adobe"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Adobe","institution_ids":["https://openalex.org/I1306409833"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031305137","display_name":"Tracy Holloway King","orcid":"https://orcid.org/0000-0002-7956-505X"},"institutions":[{"id":"https://openalex.org/I1306409833","display_name":"Adobe Systems (United States)","ror":"https://ror.org/059tvcg64","country_code":"US","type":"company","lineage":["https://openalex.org/I1306409833"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tracy Holloway King","raw_affiliation_strings":["Adobe"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Adobe","institution_ids":["https://openalex.org/I1306409833"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5128712191","display_name":"Jing Zheng","orcid":null},"institutions":[{"id":"https://openalex.org/I1306409833","display_name":"Adobe Systems (United States)","ror":"https://ror.org/059tvcg64","country_code":"US","type":"company","lineage":["https://openalex.org/I1306409833"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jing Zheng","raw_affiliation_strings":["Adobe"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Adobe","institution_ids":["https://openalex.org/I1306409833"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.67248171,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1498","last_page":"1504"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.30149999260902405,"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"}},"topics":[{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.30149999260902405,"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"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.20909999310970306,"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/T12031","display_name":"Speech and dialogue systems","score":0.04899999871850014,"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/reinforcement-learning","display_name":"Reinforcement learning","score":0.5824999809265137},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.3693000078201294},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.3569999933242798},{"id":"https://openalex.org/keywords/action","display_name":"Action (physics)","score":0.3222000002861023},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.302700012922287},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.28040000796318054}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6833000183105469},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5824999809265137},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5569999814033508},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.3693000078201294},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3569999933242798},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3506999909877777},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.3222000002861023},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.313400000333786},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.302700012922287},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.28040000796318054},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.26750001311302185},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.26440000534057617},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.2583000063896179},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.2515000104904175}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icdmw69685.2025.00179","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icdmw69685.2025.00179","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Data Mining Workshops (ICDMW)","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":1,"referenced_works":["https://openalex.org/W4387321091"],"related_works":[],"abstract_inverted_index":{"In":[0],"this":[1],"work,":[2],"we":[3],"investigate":[4],"how":[5],"small":[6],"language":[7,45,75],"models":[8,76,127],"(SLMs)":[9],"can":[10,104],"be":[11],"scaled":[12],"to":[13,49,68],"support":[14],"multimodal":[15,107,133,143],"search":[16,108],"and":[17,82,96,109,117,141,149],"recommendation":[18,110],"use":[19],"cases":[20],"while":[21,112],"remaining":[22],"efficient":[23],"enough":[24],"for":[25,61,131],"real-time,":[26],"resource-constrained":[27],"deployments.":[28],"We":[29],"present":[30],"a":[31,43,51],"framework":[32],"that":[33,102],"combines":[34],"upside-down":[35],"reinforcement":[36],"learning":[37],"with":[38],"synthetic":[39],"data":[40],"distillation":[41],"from":[42],"large":[44,74],"model":[46,59],"(Llama-3":[47],"[1])":[48],"train":[50],"<tex":[52],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[53],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$\\mathbf{1":[54],"0":[55,56],"M}$</tex>-parameter":[57],"GPT-2":[58],"[2]":[60],"multitask":[62],"prompt":[63],"generation.":[64,152],"Despite":[65],"being":[66],"up":[67],"80":[69],"times":[70],"smaller":[71],"than":[72],"state-of-the-art":[73],"(LLMs),":[77],"our":[78],"SLM":[79],"achieves":[80],"relevance":[81],"diversity":[83],"scores":[84],"within":[85],"6%":[86],"of":[87,125],"competitive":[88],"baselines":[89],"such":[90,145],"as":[91,128,146],"Llama-3":[92],"8B,":[93,95],"Qwen3":[94],"Ministral":[97],"8B.":[98],"These":[99],"results":[100],"demonstrate":[101],"SLMs":[103],"effectively":[105],"handle":[106],"tasks,":[111],"dramatically":[113],"reducing":[114],"inference":[115],"latency":[116],"memory":[118],"overhead.":[119],"Our":[120],"study":[121],"highlights":[122],"the":[123,136],"potential":[124],"lightweight":[126],"practical":[129],"engines":[130],"scalable":[132],"discovery,":[134],"bridging":[135],"gap":[137],"between":[138],"cutting-edge":[139],"research":[140],"real-world":[142],"applications":[144],"media":[147],"recommendations":[148],"creative":[150],"content":[151]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-03-12T00:00:00"}