{"id":"https://openalex.org/W4416054714","doi":"https://doi.org/10.48550/arxiv.2510.17896","title":"Long-Context Attention Benchmark: From Kernel Efficiency to Distributed Context Parallelism","display_name":"Long-Context Attention Benchmark: From Kernel Efficiency to Distributed Context Parallelism","publication_year":2025,"publication_date":"2025-10-19","ids":{"openalex":"https://openalex.org/W4416054714","doi":"https://doi.org/10.48550/arxiv.2510.17896"},"language":null,"primary_location":{"id":"pmh:oai:arXiv.org:2510.17896","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.17896","pdf_url":"https://arxiv.org/pdf/2510.17896","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"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2510.17896","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5112699885","display_name":"T. Bu","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Bu, Tao","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009961171","display_name":"Q. Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Qiangang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Zeng, Bowen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Bowen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070685695","display_name":"Hanwen Sun","orcid":"https://orcid.org/0000-0001-5093-2197"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Hanwen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113228215","display_name":"Yunpeng Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Yunpeng","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5072782399","display_name":"Chun Cao","orcid":"https://orcid.org/0000-0002-7332-073X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cao, Chun","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5080648882","display_name":"Jingwei Xu","orcid":"https://orcid.org/0000-0002-1865-6214"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Jingwei","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5112699885"],"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.17960000038146973,"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"}},"topics":[{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.17960000038146973,"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/T14347","display_name":"Big Data and Digital Economy","score":0.13989999890327454,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.12630000710487366,"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/benchmark","display_name":"Benchmark (surveying)","score":0.7343999743461609},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.6571000218391418},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.6269000172615051},{"id":"https://openalex.org/keywords/modular-design","display_name":"Modular design","score":0.5595999956130981},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4799000024795532},{"id":"https://openalex.org/keywords/kernel","display_name":"Kernel (algebra)","score":0.4738999903202057},{"id":"https://openalex.org/keywords/interface","display_name":"Interface (matter)","score":0.4302999973297119},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.3614000082015991}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8456000089645386},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.7343999743461609},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.6571000218391418},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.6269000172615051},{"id":"https://openalex.org/C101468663","wikidata":"https://www.wikidata.org/wiki/Q1620158","display_name":"Modular design","level":2,"score":0.5595999956130981},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.4803999960422516},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4799000024795532},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.4738999903202057},{"id":"https://openalex.org/C113843644","wikidata":"https://www.wikidata.org/wiki/Q901882","display_name":"Interface (matter)","level":4,"score":0.4302999973297119},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.40049999952316284},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.38179999589920044},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.37860000133514404},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.3614000082015991},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.3601999878883362},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.34940001368522644},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3474000096321106},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.34689998626708984},{"id":"https://openalex.org/C91481028","wikidata":"https://www.wikidata.org/wiki/Q1054686","display_name":"Distributed memory","level":3,"score":0.3391999900341034},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.3009999990463257},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.2849000096321106},{"id":"https://openalex.org/C122280245","wikidata":"https://www.wikidata.org/wiki/Q620622","display_name":"Kernel method","level":3,"score":0.274399995803833},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.2696000039577484},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2655999958515167},{"id":"https://openalex.org/C19768560","wikidata":"https://www.wikidata.org/wiki/Q320727","display_name":"Dependency (UML)","level":2,"score":0.26510000228881836},{"id":"https://openalex.org/C86251818","wikidata":"https://www.wikidata.org/wiki/Q816754","display_name":"Benchmarking","level":2,"score":0.26260000467300415},{"id":"https://openalex.org/C2993807640","wikidata":"https://www.wikidata.org/wiki/Q103709453","display_name":"Attention network","level":2,"score":0.25110000371932983}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2510.17896","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.17896","pdf_url":"https://arxiv.org/pdf/2510.17896","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"},{"id":"doi:10.48550/arxiv.2510.17896","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.17896","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":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2510.17896","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.17896","pdf_url":"https://arxiv.org/pdf/2510.17896","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":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Transformer-based":[0],"large":[1],"language":[2],"models":[3],"(LLMs)":[4],"have":[5],"achieved":[6],"remarkable":[7],"success,":[8],"yet":[9],"their":[10],"standard":[11],"attention":[12,48,59,66,106,129,180],"mechanism":[13],"incurs":[14],"quadratic":[15],"computation":[16],"and":[17,46,50,108,115,137,139,143,172,178],"memory":[18],"costs":[19],"with":[20,88,112],"respect":[21],"to":[22,56,161],"sequence":[23,141],"length,":[24],"posing":[25],"a":[26,100,113],"major":[27],"bottleneck":[28],"for":[29,118,176],"long-context":[30,151,183],"training.":[31,152,185],"Prior":[32],"work":[33],"tackles":[34],"this":[35],"challenge":[36],"along":[37,124],"two":[38,125],"directions:":[39],"(1)":[40,128],"kernel-level":[41],"optimizations,":[42],"which":[43,64,132,146],"accelerate":[44],"dense":[45],"sparse":[47],"operators;":[49],"(2)":[51,140],"module-level":[52],"strategies,":[53],"often":[54,79],"referred":[55],"as":[57],"distributed":[58,144],"or":[60],"context":[61,82,109],"parallel":[62,83,110],"training,":[63],"scale":[65],"across":[67,92],"multiple":[68],"devices.":[69],"However,":[70],"systematic":[71],"evaluation":[72],"still":[73],"remains":[74],"limited:":[75],"operator-level":[76],"comparisons":[77],"are":[78,85],"incomplete,":[80],"while":[81],"strategies":[84],"typically":[86],"framework-specific,":[87],"unclear":[89],"performance":[90,148],"analysis":[91],"contexts.":[93],"To":[94],"address":[95],"these":[96],"gaps,":[97],"we":[98],"propose":[99],"unified":[101],"benchmark":[102,121,165],"that":[103],"integrates":[104],"representative":[105],"kernels":[107],"mechanisms":[111,181],"modular":[114],"extensible":[116],"interface":[117],"evaluation.":[119],"The":[120],"evaluates":[122],"methods":[123],"critical":[126],"dimensions:":[127],"mask":[130],"patterns,":[131],"strongly":[133],"affect":[134],"efficiency,":[135],"scalability,":[136],"usability,":[138],"length":[142],"scale,":[145],"determine":[147],"under":[149],"extreme":[150],"Through":[153],"comprehensive":[154],"experiments":[155],"on":[156],"the":[157],"cluster":[158],"of":[159],"up":[160],"96":[162],"GPUs,":[163],"our":[164],"enables":[166],"reproducible":[167],"comparisons,":[168],"highlights":[169],"method-specific":[170],"trade-offs,":[171],"provides":[173],"practical":[174],"guidance":[175],"designing":[177],"deploying":[179],"in":[182],"LLM":[184]},"counts_by_year":[],"updated_date":"2026-04-21T08:09:41.155169","created_date":"2025-10-24T00:00:00"}