{"id":"https://openalex.org/W4412888721","doi":"https://doi.org/10.18653/v1/2025.findings-acl.131","title":"RaaS: Reasoning-Aware Attention Sparsity for Efficient LLM Reasoning","display_name":"RaaS: Reasoning-Aware Attention Sparsity for Efficient LLM Reasoning","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888721","doi":"https://doi.org/10.18653/v1/2025.findings-acl.131"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.131","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.131","pdf_url":"https://aclanthology.org/2025.findings-acl.131.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.131.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5004435008","display_name":"Junhao Hu","orcid":"https://orcid.org/0000-0003-0699-1119"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Junhao Hu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101823721","display_name":"Wenrui Huang","orcid":"https://orcid.org/0009-0005-8881-5005"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wenrui Huang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100432932","display_name":"Weidong Wang","orcid":"https://orcid.org/0000-0002-3550-0625"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Weidong Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5061052661","display_name":"Zhenwen Li","orcid":"https://orcid.org/0009-0000-3209-0272"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhenwen Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006591088","display_name":"Tiancheng Hu","orcid":"https://orcid.org/0009-0006-7354-1088"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tiancheng Hu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035543428","display_name":"Zhixia Liu","orcid":"https://orcid.org/0000-0001-5605-9747"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhixia Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101527007","display_name":"Xusheng Chen","orcid":"https://orcid.org/0000-0002-2807-9780"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xusheng Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012942103","display_name":"Tao Xie","orcid":"https://orcid.org/0000-0002-7489-6861"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tao Xie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5060703788","display_name":"Yizhou Shan","orcid":"https://orcid.org/0009-0007-9519-0546"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yizhou Shan","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":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"2577","last_page":"2590"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10215","display_name":"Semantic Web and Ontologies","score":0.9764999747276306,"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/T10215","display_name":"Semantic Web and Ontologies","score":0.9764999747276306,"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/T10703","display_name":"Business Process Modeling and Analysis","score":0.942300021648407,"subfield":{"id":"https://openalex.org/subfields/1404","display_name":"Management Information Systems"},"field":{"id":"https://openalex.org/fields/14","display_name":"Business, Management and Accounting"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T11063","display_name":"Rough Sets and Fuzzy Logic","score":0.9315999746322632,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/computer-science","display_name":"Computer science","score":0.651073694229126},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.34072959423065186}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.651073694229126},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.34072959423065186}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.131","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.131","pdf_url":"https://aclanthology.org/2025.findings-acl.131.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.131","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.131","pdf_url":"https://aclanthology.org/2025.findings-acl.131.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G6317451256","display_name":null,"funder_award_id":"92464301","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/W4412888721.pdf","grobid_xml":"https://content.openalex.org/works/W4412888721.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4],"demonstrated":[5],"strong":[6],"capabilities":[7],"across":[8],"various":[9],"domains,":[10],"with":[11,73,90,169],"recent":[12],"advancements":[13],"in":[14,109,131],"challenging":[15],"reasoning":[16,24,123],"tasks":[17,25],"such":[18],"as":[19],"mathematics":[20],"and":[21,38,80,137,155,172],"programming.However,":[22],"solving":[23],"often":[26],"requires":[27],"an":[28],"LLM":[29],"to":[30,56,129],"generate":[31],"long":[32],"sequences,":[33],"incurring":[34],"O(N":[35,94],")":[36,95],"time":[37,92,171],"memory":[39,96,174],"complexities":[40],"per":[41],"token,":[42],"where":[43,125],"N":[44,103],"is":[45,98],"the":[46,61,66,74,83,99,106,119],"current":[47],"sequence":[48],"length.To":[49],"reduce":[50],"complexities,":[51],"existing":[52],"sparsity-based":[53],"algorithms":[54,71],"propose":[55,146],"retain":[57],"Key-Value":[58],"(KV)":[59],"vectors,":[60],"intermediate":[62],"representations":[63],"of":[64,77,122],"only":[65],"most":[67],"critical":[68],"tokens.However,":[69],"these":[70],"struggle":[72],"\"impossible":[75,107],"trinity\"":[76],"accuracy,":[78],"time,":[79],"memory.For":[81],"example,":[82],"state-of-the-art":[84],"algorithm,":[85],"Quest,":[86],"achieves":[87],"high":[88,167],"accuracy":[89,168],"O(L)":[91,170,173],"but":[93],"(L":[97],"cache":[100],"budget,":[101],"L":[102],").To":[104],"address":[105],"trinity\",":[108],"this":[110,143],"paper,":[111],"we":[112,145],"identify":[113],"a":[114,147],"new":[115,148],"attention":[116],"pattern":[117],"during":[118],"decode":[120],"stage":[121],"tasks,":[124],"milestone":[126,153],"tokens":[127,154],"(analogous":[128],"lemmas":[130],"mathematical":[132],"proofs)":[133],"emerge,":[134],"are":[135,162],"utilized,":[136],"then":[138],"become":[139],"unimportant":[140],"afterward.Based":[141],"on":[142],"pattern,":[144],"algorithm":[149],"RaaS":[150],"that":[151],"identifies":[152],"retains":[156],"their":[157],"KV":[158],"vectors":[159],"until":[160],"they":[161],"no":[163],"longer":[164],"needed,":[165],"achieving":[166],"complexities.":[175]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}