{"id":"https://openalex.org/W7224837786","doi":"https://doi.org/10.48550/arxiv.2610.37973","title":"Merge and Branch: Efficient Tree-Structured Supervised Fine-Tuning of LLMs for Recommendation","display_name":"Merge and Branch: Efficient Tree-Structured Supervised Fine-Tuning of LLMs for Recommendation","publication_year":2026,"publication_date":"2026-10-27","ids":{"openalex":"https://openalex.org/W7224837786","doi":"https://doi.org/10.48550/arxiv.2610.37973"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.37973","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.37973","pdf_url":"https://arxiv.org/pdf/2610.37973","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.37973","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101268145","display_name":"Yunqin Zhu","orcid":"https://orcid.org/0009-0002-9709-2417","observed_orcids":["https://orcid.org/0009-0002-9709-2417"]},"institutions":[{"id":"https://openalex.org/I130701444","display_name":"Georgia Institute of Technology","ror":"https://ror.org/01zkghx44","country_code":"US","type":"education","lineage":["https://openalex.org/I130701444"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhu, Yunqin","raw_affiliation_strings":["Georgia Institute of Technology Atlanta , GA , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Georgia Institute of Technology Atlanta , GA , USA","institution_ids":["https://openalex.org/I130701444"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125616570","display_name":"Liyang Shu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shu, Liyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5161097540","display_name":"Yusi Zhang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yusi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160207755","display_name":"Yao Xie","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I130701444","display_name":"Georgia Institute of Technology","ror":"https://ror.org/01zkghx44","country_code":"US","type":"education","lineage":["https://openalex.org/I130701444"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Xie, Yao","raw_affiliation_strings":["Georgia Institute of Technology Atlanta , GA , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Georgia Institute of Technology Atlanta , GA , USA","institution_ids":["https://openalex.org/I130701444"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5161100016","display_name":"Tianhao Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Tianhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5055504013","display_name":"Xinhang Song","orcid":"https://orcid.org/0000-0002-0895-1076","observed_orcids":["https://orcid.org/0000-0002-0895-1076"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Song, Xinhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5161140164","display_name":"Tianyu Wen","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wen, Tianyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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":0.9994999766349792,"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":0.9994999766349792,"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/T10286","display_name":"Information Retrieval and Search Behavior","score":9.999999747378752e-05,"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":0.0,"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/llm-based-recommendation","display_name":"LLM-based recommendation","score":0.8259999752044678},{"id":"https://openalex.org/keywords/supervised-fine-tuning","display_name":"supervised fine-tuning","score":0.8690000176429749},{"id":"https://openalex.org/keywords/flashattention","display_name":"FlashAttention","score":0.5559999942779541},{"id":"https://openalex.org/keywords/key-value-cache","display_name":"key-value cache","score":0.8550000190734863},{"id":"https://openalex.org/keywords/generative-recommendation","display_name":"generative recommendation","score":0.5950000286102295},{"id":"https://openalex.org/keywords/e-commerce-recommender-systems","display_name":"e-commerce recommendation","score":0.7170000076293945},{"id":"https://openalex.org/keywords/training-efficiency","display_name":"training efficiency","score":0.5830000042915344},{"id":"https://openalex.org/keywords/constrained-decoding","display_name":"constrained decoding","score":0.9190000295639038}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7745000123977661},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.7189000248908997},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.6200000047683716},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.4496999979019165},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.43529999256134033},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41449999809265137},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.3939000070095062},{"id":"https://openalex.org/C2776636908","wikidata":"https://www.wikidata.org/wiki/Q441989","display_name":"Public history","level":2,"score":0.38089999556541443},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.36550000309944153},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.3653999865055084},{"id":"https://openalex.org/C140086265","wikidata":"https://www.wikidata.org/wiki/Q11341754","display_name":"Merge algorithm","level":4,"score":0.32580000162124634},{"id":"https://openalex.org/C154504017","wikidata":"https://www.wikidata.org/wiki/Q853614","display_name":"Identifier","level":2,"score":0.31529998779296875},{"id":"https://openalex.org/C119839945","wikidata":"https://www.wikidata.org/wiki/Q6545185","display_name":"Unique identifier","level":3,"score":0.3050999939441681},{"id":"https://openalex.org/C2775945657","wikidata":"https://www.wikidata.org/wiki/Q381442","display_name":"Structuring","level":2,"score":0.2858000099658966},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.27649998664855957},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.2718000113964081},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.27070000767707825},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.260699987411499}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.37973","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.37973","pdf_url":"https://arxiv.org/pdf/2610.37973","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2610.37973","doi":"https://doi.org/10.48550/arxiv.2610.37973","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.37973","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,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2610.37973","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.37973","pdf_url":"https://arxiv.org/pdf/2610.37973","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7224837786.pdf","grobid_xml":"https://content.openalex.org/works/W7224837786.grobid-xml"},"referenced_works_count":34,"referenced_works":["https://openalex.org/W2963367478","https://openalex.org/W2984100107","https://openalex.org/W4296591867","https://openalex.org/W4297971002","https://openalex.org/W4376312036","https://openalex.org/W4384644330","https://openalex.org/W4386728933","https://openalex.org/W4386729798","https://openalex.org/W4386730465","https://openalex.org/W4390412407","https://openalex.org/W4395112660","https://openalex.org/W4400525124","https://openalex.org/W4400532459","https://openalex.org/W4400909953","https://openalex.org/W4401863330","https://openalex.org/W4402671542","https://openalex.org/W4403577940","https://openalex.org/W4407375814","https://openalex.org/W4407403520","https://openalex.org/W4407814501","https://openalex.org/W4412376956","https://openalex.org/W4412394905","https://openalex.org/W4412825745","https://openalex.org/W4414034912","https://openalex.org/W4415797413","https://openalex.org/W4415937555","https://openalex.org/W4416043540","https://openalex.org/W7133227460","https://openalex.org/W7133233095","https://openalex.org/W7133239020","https://openalex.org/W7156045069","https://openalex.org/W7165366801","https://openalex.org/W7167614485","https://openalex.org/W7168406275"],"related_works":["https://openalex.org/W7171635594","https://openalex.org/W4412377020","https://openalex.org/W7169854116","https://openalex.org/W4412876835","https://openalex.org/W4415937555","https://openalex.org/W4417472355","https://openalex.org/W7223660444","https://openalex.org/W7160828677","https://openalex.org/W7221761071","https://openalex.org/W7204882631"],"abstract_inverted_index":{"LLM-based":[0],"generative":[1],"recommendation":[2,138],"reads":[3],"long,":[4],"text-rich":[5],"interaction":[6],"histories":[7],"but":[8],"generates":[9],"only":[10,33],"short":[11],"semantic":[12],"identifiers":[13],"(SIDs).":[14],"This":[15],"asymmetry":[16],"makes":[17],"supervised":[18],"fine-tuning":[19],"inefficient:":[20],"encoding":[21],"the":[22,70,106],"history":[23,28,72,80,108,124],"is":[24],"expensive,":[25],"yet":[26],"each":[27,50,55,89],"usually":[29],"provides":[30],"supervision":[31],"for":[32,116],"one":[34],"target.":[35],"We":[36,96],"present":[37],"Merge":[38,53],"and":[39,46,59,110,144,173],"Branch":[40,68],"(M&B)":[41],"to":[42,73,100,126],"reduce":[43],"input":[44],"cost":[45],"learn":[47],"more":[48],"from":[49,113],"encoded":[51],"history.":[52],"compresses":[54],"historical":[56],"item's":[57],"SID":[58],"text":[60,161],"into":[61],"a":[62,92,151],"small":[63],"number":[64],"of":[65],"learned":[66],"tokens.":[67],"uses":[69],"compressed":[71],"predict":[74],"multiple":[75],"target":[76],"SIDs.":[77],"To":[78],"share":[79],"computation":[81],"while":[82,140],"keeping":[83],"these":[84],"targets":[85],"independent,":[86],"M&B":[87,135,157],"organizes":[88],"example":[90],"as":[91],"shallow,":[93],"wide":[94],"tree.":[95],"develop":[97],"tree":[98],"FlashAttention":[99],"execute":[101],"this":[102],"layout":[103],"efficiently,":[104],"processing":[105],"shared":[107],"once":[109],"propagating":[111],"gradients":[112],"every":[114],"branch":[115],"end-to-end":[117],"training.":[118],"A":[119],"tree-structured":[120],"key/value":[121],"cache":[122],"extends":[123],"sharing":[125],"constrained":[127],"decoding.":[128],"Experiments":[129],"on":[130,150],"public":[131],"benchmarks":[132],"show":[133],"that":[134,156],"outperforms":[136],"strong":[137],"baselines":[139],"substantially":[141],"improving":[142],"training":[143],"decoding":[145],"efficiency.":[146],"Online":[147],"A/B":[148],"tests":[149],"large-scale":[152],"e-commerce":[153],"platform":[154],"suggest":[155],"effectively":[158],"combines":[159],"item":[160],"with":[162],"multi-target":[163],"supervision,":[164],"yielding":[165],"statistically":[166],"significant":[167],"gains":[168],"in":[169],"per-user":[170],"clicks,":[171],"orders,":[172],"GMV.":[174]},"counts_by_year":[],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2026-10-31T00:00:00"}