{"id":"https://openalex.org/W7166880451","doi":"https://doi.org/10.18653/v1/2026.acl-long.51","title":"Exploring Attention Attractors in Large Language Models","display_name":"Exploring Attention Attractors in Large Language Models","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166880451","doi":"https://doi.org/10.18653/v1/2026.acl-long.51"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.acl-long.51","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.51","pdf_url":"https://aclanthology.org/2026.acl-long.51.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.51.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139799115","display_name":"Ziheng Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I78988378","display_name":"Renmin University of China","ror":"https://ror.org/041pakw92","country_code":"CN","type":"education","lineage":["https://openalex.org/I78988378"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ziheng Wang","raw_affiliation_strings":["AIM3 Lab , Renmin University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AIM3 Lab , Renmin University of China","institution_ids":["https://openalex.org/I78988378"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104436927","display_name":"Zihao Yue","orcid":"https://orcid.org/0000-0002-3470-5442"},"institutions":[{"id":"https://openalex.org/I78988378","display_name":"Renmin University of China","ror":"https://ror.org/041pakw92","country_code":"CN","type":"education","lineage":["https://openalex.org/I78988378"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zihao Yue","raw_affiliation_strings":["AIM3 Lab , Renmin University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AIM3 Lab , Renmin University of China","institution_ids":["https://openalex.org/I78988378"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139837199","display_name":"Wenxuan Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I78988378","display_name":"Renmin University of China","ror":"https://ror.org/041pakw92","country_code":"CN","type":"education","lineage":["https://openalex.org/I78988378"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wenxuan Wang","raw_affiliation_strings":["AIM3 Lab , Renmin University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AIM3 Lab , Renmin University of China","institution_ids":["https://openalex.org/I78988378"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5139794858","display_name":"Qin Jin","orcid":null},"institutions":[{"id":"https://openalex.org/I78988378","display_name":"Renmin University of China","ror":"https://ror.org/041pakw92","country_code":"CN","type":"education","lineage":["https://openalex.org/I78988378"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qin Jin","raw_affiliation_strings":["AIM3 Lab , Renmin University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AIM3 Lab , Renmin University of China","institution_ids":["https://openalex.org/I78988378"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I78988378"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.86873604,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1148","last_page":"1160"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.19480000436306,"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":0.19480000436306,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.08550000190734863,"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"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.06790000200271606,"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/language-model","display_name":"Language model","score":0.35350000858306885},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.2928999960422516},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.28299999237060547},{"id":"https://openalex.org/keywords/attractor","display_name":"Attractor","score":0.2689000070095062},{"id":"https://openalex.org/keywords/domain","display_name":"Domain (mathematical analysis)","score":0.26820001006126404}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5511000156402588},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.42579999566078186},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.38089999556541443},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.35350000858306885},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.35339999198913574},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.3068000078201294},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2928999960422516},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.28299999237060547},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.2777000069618225},{"id":"https://openalex.org/C164380108","wikidata":"https://www.wikidata.org/wiki/Q507187","display_name":"Attractor","level":2,"score":0.2689000070095062},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.26820001006126404},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.25619998574256897},{"id":"https://openalex.org/C511192102","wikidata":"https://www.wikidata.org/wiki/Q5156948","display_name":"Comprehension","level":2,"score":0.25380000472068787}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.51","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.51","pdf_url":"https://aclanthology.org/2026.acl-long.51.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.51","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.51","pdf_url":"https://aclanthology.org/2026.acl-long.51.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":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.5296764373779297}],"awards":[],"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/W7166880451.pdf","grobid_xml":"https://content.openalex.org/works/W7166880451.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"This":[0],"paper":[1],"explores":[2],"attention":[3,43,64,83],"attractorstokens":[4],"that":[5,42],"draw":[6],"significantly":[7],"high":[8],"attentionin":[9],"large":[10,86],"language":[11,87],"models.We":[12],"analyze":[13],"them":[14],"from":[15,27,53],"three":[16],"perspectives:":[17],"(1)":[18],"Functionality:":[19],"We":[20,59],"demonstrate":[21,60],"their":[22,70],"role":[23],"in":[24],"aggregating":[25],"information":[26],"preceding":[28],"contexts":[29],"to":[30,67],"facilitate":[31],"future":[32],"predictions.(2)":[33],"Distribution:":[34],"Through":[35],"layer-wise":[36],"and":[37,89],"token-wise":[38],"analysis,":[39],"we":[40],"reveal":[41],"attractors":[44],"are":[45],"widely":[46],"distributed":[47],"across":[48],"layers":[49],"but":[50],"predominantly":[51],"originate":[52],"low-semantic":[54],"words":[55],"like":[56],"\"_the\".(3)":[57],"Mechanism:":[58],"the":[61,82],"correlation":[62],"between":[63],"weights":[65],"allocated":[66],"tokens":[68],"with":[69],"specific":[71],"activation":[72],"dimension":[73],"values.We":[74],"hope":[75],"these":[76],"findings":[77],"provide":[78],"new":[79],"insights":[80],"into":[81],"mechanisms":[84],"of":[85],"models":[88],"inspire":[90],"further":[91],"exploration.":[92]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}