{"id":"https://openalex.org/W7220649409","doi":"https://doi.org/10.48550/arxiv.2610.05559","title":"Cut Binary Cross Entropy: Efficient Large-Vocabulary Loss and Gradient Kernels for Sequential Recommendation","display_name":"Cut Binary Cross Entropy: Efficient Large-Vocabulary Loss and Gradient Kernels for Sequential Recommendation","publication_year":2026,"publication_date":"2026-10-04","ids":{"openalex":"https://openalex.org/W7220649409","doi":"https://doi.org/10.48550/arxiv.2610.05559"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.05559","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05559","pdf_url":"https://arxiv.org/pdf/2610.05559","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":"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/2610.05559","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5002228588","display_name":"Yaoyiran Li","orcid":"https://orcid.org/0000-0001-9529-3878","observed_orcids":["https://orcid.org/0000-0001-9529-3878"]},"institutions":[{"id":"https://openalex.org/I4210113297","display_name":"Google (United Kingdom)","ror":"https://ror.org/024bc3e07","country_code":"GB","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210113297","https://openalex.org/I4210128969"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Li, Yaoyiran","raw_affiliation_strings":["Google Cloud London , United Kingdom"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Cloud London , United Kingdom","institution_ids":["https://openalex.org/I4210113297"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5157729133","display_name":"Haowen Ning","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ning, Haowen","raw_affiliation_strings":["Google Cloud Mountain View , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Cloud Mountain View , USA","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5135855381","display_name":"Mohamed Hammad","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hammad, Mohamed","raw_affiliation_strings":["Google Cloud Mountain View , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Cloud Mountain View , USA","institution_ids":["https://openalex.org/I1291425158"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"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":0.8407999873161316,"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":0.8407999873161316,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.13289999961853027,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10036","display_name":"Advanced Neural Network Applications","score":0.010599999688565731,"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/binary-cross-entropy-loss","display_name":"binary cross-entropy loss","score":0.8610000014305115},{"id":"https://openalex.org/keywords/sequential-recommendation","display_name":"sequential recommendation","score":0.8700000047683716},{"id":"https://openalex.org/keywords/out-of-memory-errors","display_name":"out-of-memory errors","score":0.7480000257492065},{"id":"https://openalex.org/keywords/high-bandwidth-memory","display_name":"high-bandwidth memory","score":0.7400000095367432},{"id":"https://openalex.org/keywords/tpu","display_name":"TPU","score":0.5299999713897705},{"id":"https://openalex.org/keywords/jax","display_name":"JAX","score":0.49799999594688416},{"id":"https://openalex.org/keywords/pallas","display_name":"Pallas","score":0.8799999952316284},{"id":"https://openalex.org/keywords/memory-efficient-training","display_name":"memory-efficient training","score":0.45100000500679016},{"id":"https://openalex.org/keywords/distributed-training","display_name":"distributed training","score":0.4580000042915344},{"id":"https://openalex.org/keywords/sasrec","display_name":"SASRec","score":0.5189999938011169}],"concepts":[{"id":"https://openalex.org/C188441871","wikidata":"https://www.wikidata.org/wiki/Q7554146","display_name":"Softmax function","level":3,"score":0.7121999859809875},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6474999785423279},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.6355999708175659},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.5957000255584717},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5523999929428101},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4478999972343445},{"id":"https://openalex.org/C17020691","wikidata":"https://www.wikidata.org/wiki/Q139677","display_name":"Operator (biology)","level":5,"score":0.414000004529953},{"id":"https://openalex.org/C140331021","wikidata":"https://www.wikidata.org/wiki/Q1868104","display_name":"Logit","level":2,"score":0.3828999996185303},{"id":"https://openalex.org/C90673727","wikidata":"https://www.wikidata.org/wiki/Q901718","display_name":"Product (mathematics)","level":2,"score":0.37119999527931213},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.36070001125335693},{"id":"https://openalex.org/C180623205","wikidata":"https://www.wikidata.org/wiki/Q1268589","display_name":"Outer product","level":3,"score":0.35740000009536743},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.3547999858856201},{"id":"https://openalex.org/C2776257435","wikidata":"https://www.wikidata.org/wiki/Q1576430","display_name":"Bandwidth (computing)","level":2,"score":0.33550000190734863},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3310999870300293},{"id":"https://openalex.org/C66905080","wikidata":"https://www.wikidata.org/wiki/Q17005494","display_name":"Binary classification","level":3,"score":0.3052000105381012},{"id":"https://openalex.org/C557471498","wikidata":"https://www.wikidata.org/wiki/Q554950","display_name":"Recommender system","level":2,"score":0.2939999997615814},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2743000090122223},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2603999972343445},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.2574000060558319},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2565999925136566}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.05559","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05559","pdf_url":"https://arxiv.org/pdf/2610.05559","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2610.05559","doi":"https://doi.org/10.48550/arxiv.2610.05559","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.05559","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:2610.05559","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05559","pdf_url":"https://arxiv.org/pdf/2610.05559","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":"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/W7220649409.pdf","grobid_xml":"https://content.openalex.org/works/W7220649409.grobid-xml"},"referenced_works_count":3,"referenced_works":["https://openalex.org/W4403573100","https://openalex.org/W4407632696","https://openalex.org/W4414887411"],"related_works":["https://openalex.org/W4403582629","https://openalex.org/W4414591557","https://openalex.org/W4394840362","https://openalex.org/W4403220659","https://openalex.org/W4404449926","https://openalex.org/W4321636622","https://openalex.org/W7090908489","https://openalex.org/W7202357449","https://openalex.org/W7157258572","https://openalex.org/W3121395088"],"abstract_inverted_index":{"Industrial":[0],"sequential":[1],"recommender":[2],"systems":[3],"operate":[4],"over":[5,22],"massive":[6],"item":[7],"catalogs":[8],"(e.g.,":[9],"10^5--10^7":[10],"items).":[11],"Multi-label":[12],"recommendation":[13],"models":[14],"are":[15],"trained":[16],"with":[17,111,160,173,194],"Binary":[18],"Cross-Entropy":[19,58],"(BCE)":[20],"loss":[21,53,78,100],"the":[23],"full":[24],"vocabulary,":[25],"but":[26],"standard":[27],"BCE":[28,63,77],"materializes":[29],"a":[30,106,112],"dense":[31,98],"[B,":[32],"N,":[33],"V]":[34],"logits":[35,126],"tensor":[36],"in":[37,59,83,122,132],"High":[38],"Bandwidth":[39],"Memory":[40],"(HBM),":[41],"incurring":[42],"prohibitive":[43],"$O(BNV)$":[44],"memory":[45],"and":[46,79,85,101,127,138,145,188],"fatal":[47],"Out-Of-Memory":[48],"(OOM)":[49],"errors.":[50],"While":[51],"chunked":[52],"optimizations":[54],"exist":[55],"for":[56,87,142,170],"Softmax":[57],"LLMs,":[60],"large-scale":[61],"multi-label":[62,171],"optimization":[64],"remains":[65],"unexplored":[66],"across":[67],"deep":[68],"learning":[69],"ecosystems.":[70],"We":[71],"propose":[72],"CutBCE,":[73],"an":[74,93],"exact,":[75],"hardware-accelerated":[76],"gradient":[80],"operator":[81],"implemented":[82],"JAX":[84],"Pallas":[86,114],"large-vocabulary":[88],"workloads.":[89],"CutBCE":[90,156,177,197],"introduces":[91],"(1)":[92],"exact":[94],"fused":[95],"reformulation":[96],"evaluating":[97],"background":[99],"sparse":[102],"target":[103],"corrections;":[104],"(2)":[105],"custom":[107],"Vector-Jacobian":[108],"Product":[109],"(VJP)":[110],"dedicated":[113],"TPU":[115,153,167],"backward":[116],"kernel":[117],"computing":[118],"logit":[119],"tiles":[120],"on-chip":[121],"both":[123],"passes":[124],"so":[125],"their":[128],"gradients":[129],"never":[130],"reside":[131],"HBM;":[133],"(3)":[134],"dynamic":[135],"VMEM":[136],"budgeting":[137],"sharding-aware":[139],"collective":[140],"hoisting":[141],"distributed":[143],"meshes;":[144],"(4)":[146],"count-based":[147],"zero-overhead":[148],"training":[149,169,190],"metrics.":[150],"On":[151,165],"single-chip":[152],"v5e/v6e":[154],"mini-benchmarks,":[155],"eliminates":[157],"OOM":[158],"errors":[159],"up":[161],"to":[162],"91.9%":[163],"speedup.":[164],"8-chip":[166],"slice":[168],"SASRec":[172],"876k":[174],"items":[175],"(Yambda-50M),":[176],"reduces":[178],"peak":[179],"HBM":[180],"by":[181,192],"65.7%":[182],"(>14":[183],"GiB":[184],"saved":[185],"per":[186],"chip)":[187],"increases":[189],"speed":[191],"225.9%":[193],"comparable":[195],"accuracy.":[196],"is":[198],"open-sourced":[199],"at":[200],"https://github.com/AI-Hypercomputer/RecML/blob/main/recml/core/ops/binary_cross_entropy_ops.py.":[201]},"counts_by_year":[],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2026-10-07T00:00:00"}