{"id":"https://openalex.org/W7166827041","doi":"https://doi.org/10.18653/v1/2026.acl-long.46","title":"RoBSA: RoPE-based Blockwise Sparse Multi-head Latent Attention","display_name":"RoBSA: RoPE-based Blockwise Sparse Multi-head Latent Attention","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166827041","doi":"https://doi.org/10.18653/v1/2026.acl-long.46"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.acl-long.46","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.46","pdf_url":"https://aclanthology.org/2026.acl-long.46.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.acl-long.46.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139813115","display_name":"Xinyu Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinyu Shi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5044575083","display_name":"Kairong Luo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kairong Luo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139762603","display_name":"Zhen Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhen Zheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5103141832","display_name":"Wenguang Chen","orcid":"https://orcid.org/0000-0002-4281-1018"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wenguang Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.81662943,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1028","last_page":"1044"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.2401999980211258,"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/T10028","display_name":"Topic Modeling","score":0.2401999980211258,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.10939999669790268,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.10490000247955322,"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/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.41290000081062317},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.3089999854564667},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.3010999858379364},{"id":"https://openalex.org/keywords/sequence","display_name":"Sequence (biology)","score":0.28139999508857727},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.2484000027179718}],"concepts":[{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.583899974822998},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5807999968528748},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.41290000081062317},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3089999854564667},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.3010999858379364},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.28139999508857727},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2484000027179718},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.24549999833106995},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.23149999976158142},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.22679999470710754}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.46","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.46","pdf_url":"https://aclanthology.org/2026.acl-long.46.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.46","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.46","pdf_url":"https://aclanthology.org/2026.acl-long.46.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G8141470120","display_name":null,"funder_award_id":"62495062","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166827041.pdf","grobid_xml":"https://content.openalex.org/works/W7166827041.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4],"rapidly":[5],"advanced":[6],"in":[7,12,42,125,140,150,166],"recent":[8],"years,":[9],"scaling":[10],"up":[11,155],"both":[13],"parameter":[14],"count":[15],"and":[16,40,49,133],"context":[17,20,71],"length.However,":[18],"as":[19,57],"windows":[21],"extend":[22],"from":[23],"thousands":[24,28],"to":[25,121,156,163],"hundreds":[26],"of":[27,29,37,51,70,85,110,117],"tokens,":[30],"attention":[31,54,65,75,165],"computation":[32],"becomes":[33],"the":[34,47,73,107,114,151],"dominant":[35],"source":[36],"memory":[38],"usage":[39],"runtime":[41],"decoding":[43,108,152],"stages,":[44],"severely":[45],"limiting":[46],"efficiency":[48],"scalability":[50],"longcontext":[52],"LLMs.Sparse":[53],"has":[55],"emerged":[56],"a":[58,68,83,100,126,130,141],"promising":[59],"solution,":[60],"reducing":[61],"complexity":[62],"by":[63,154],"computing":[64],"over":[66],"only":[67],"subset":[69],"tokens.However,":[72],"sparse":[74],"for":[76,104,169],"Multi-head":[77],"Latent":[78],"Attention":[79,98],"(MLA)":[80],"which":[81],"is":[82,88,129],"variant":[84],"standard":[86],"MHA":[87],"rarely":[89],"studied.In":[90],"this":[91],"paper,":[92],"we":[93],"introduce":[94],"RoPE-based":[95],"Blockwise":[96],"Sparse":[97],"(RoBSA),":[99],"method":[101,144],"designed":[102],"specifically":[103],"MLA":[105,120],"during":[106],"stage":[109,153],"model":[111],"inference.RoBSA":[112],"leverages":[113],"decoupled":[115],"nature":[116],"RoPE":[118],"within":[119],"implement":[122],"token":[123],"selection":[124],"blockwise":[127],"manner.RoBSA":[128],"lightweight,":[131],"training-free,":[132],"layer-aware":[134],"algorithm":[135],"that":[136],"can":[137],"be":[138],"integrated":[139],"plug-and-play":[142],"fashion.Our":[143],"significantly":[145],"reduces":[146],"end-to-end":[147],"inference":[148],"latency":[149],"2.55\u00d7":[157],"with":[158],"minimal":[159],"accuracy":[160],"loss":[161],"compared":[162],"full":[164],"long-context":[167],"scenarios":[168],"very":[170],"large":[171],"models.":[172]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}