{"id":"https://openalex.org/W7166891261","doi":"https://doi.org/10.18653/v1/2026.findings-acl.323","title":"One Token Is Enough: Improving Diffusion Language Models with a Sink Token","display_name":"One Token Is Enough: Improving Diffusion Language Models with a Sink Token","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166891261","doi":"https://doi.org/10.18653/v1/2026.findings-acl.323"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.323","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.323","pdf_url":"https://aclanthology.org/2026.findings-acl.323.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 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.323.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5040834836","display_name":"Zihou Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zihou Zhang","raw_affiliation_strings":["University of Chinese Academy of Sciences","Xiaohongshu Inc"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210165038"]},{"raw_affiliation_string":"Xiaohongshu Inc","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139796020","display_name":"Zheyong Xie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheyong Xie","raw_affiliation_strings":["Xiaohongshu Inc"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Xiaohongshu Inc","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124107280","display_name":"Li Zhong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li Zhong","raw_affiliation_strings":["Xiaohongshu Inc"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Xiaohongshu Inc","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139751447","display_name":"Haifeng Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haifeng Liu","raw_affiliation_strings":["Xiaohongshu Inc"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Xiaohongshu Inc","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5139797858","display_name":"Shaosheng Cao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shaosheng Cao","raw_affiliation_strings":["Xiaohongshu Inc"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Xiaohongshu Inc","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"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":0.84677792,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"6479","last_page":"6490"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.48100000619888306,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.48100000619888306,"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/T10028","display_name":"Topic Modeling","score":0.11590000241994858,"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/T12031","display_name":"Speech and dialogue systems","score":0.05339999869465828,"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/security-token","display_name":"Security token","score":0.6384999752044678},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.34040001034736633},{"id":"https://openalex.org/keywords/token-passing","display_name":"Token passing","score":0.321399986743927},{"id":"https://openalex.org/keywords/modeling-language","display_name":"Modeling language","score":0.2824000120162964},{"id":"https://openalex.org/keywords/diffusion","display_name":"Diffusion","score":0.2750000059604645}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6995000243186951},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6384999752044678},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.4041000008583069},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.37959998846054077},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.34040001034736633},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3377000093460083},{"id":"https://openalex.org/C115067241","wikidata":"https://www.wikidata.org/wiki/Q1639854","display_name":"Token passing","level":3,"score":0.321399986743927},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.31439998745918274},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.2955000102519989},{"id":"https://openalex.org/C179603123","wikidata":"https://www.wikidata.org/wiki/Q1941921","display_name":"Modeling language","level":3,"score":0.2824000120162964},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.2750000059604645},{"id":"https://openalex.org/C141218545","wikidata":"https://www.wikidata.org/wiki/Q7521336","display_name":"Simulation language","level":2,"score":0.265500009059906}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.323","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.323","pdf_url":"https://aclanthology.org/2026.findings-acl.323.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 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.323","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.323","pdf_url":"https://aclanthology.org/2026.findings-acl.323.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 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.40356749296188354}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166891261.pdf","grobid_xml":"https://content.openalex.org/works/W7166891261.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Diffusion":[0],"Language":[1],"Models":[2],"(DLMs)":[3],"have":[4],"emerged":[5],"as":[6,53,147],"a":[7,24,54,77,86,92,113,148],"compelling":[8],"alternative":[9],"to":[10,59,96,99,105],"autoregressive":[11],"approaches,":[12],"enabling":[13],"parallel":[14],"text":[15],"generation":[16],"with":[17],"competitive":[18],"performance.Despite":[19],"these":[20],"advantages,":[21],"there":[22],"is":[23,133,155],"critical":[25],"instability":[26],"in":[27,41,57],"DLMs:":[28],"the":[29,42,48,128],"moving":[30,49],"sink":[31,36,50,82],"phenomenon.Our":[32],"analysis":[33,125],"indicates":[34],"that":[35,47,111,127],"tokens":[37],"exhibit":[38],"low-norm":[39],"representations":[40],"Transformer's":[43],"value":[44],"space,":[45],"and":[46,138,150],"phenomenon":[51],"serves":[52],"protective":[55],"mechanism":[56],"DLMs":[58],"prevent":[60],"excessive":[61],"information":[62],"mixing.However,":[63],"their":[64],"unpredictable":[65],"positions":[66],"across":[67],"diffusion":[68],"steps":[69],"undermine":[70],"inference":[71],"robustness.To":[72],"resolve":[73],"this,":[74],"we":[75,90],"propose":[76],"simple":[78],"but":[79],"effective":[80],"extra":[81,115],"token":[83,94,116,132],"implemented":[84],"via":[85],"modified":[87],"attention":[88,118],"mask.Specifically,":[89],"introduce":[91],"special":[93],"constrained":[95],"attend":[97],"solely":[98],"itself,":[100],"while":[101],"remaining":[102],"globally":[103],"visible":[104],"all":[106],"other":[107],"tokens.Experimental":[108],"results":[109],"demonstrate":[110],"introducing":[112],"single":[114],"stabilizes":[117],"sinks,":[119],"substantially":[120],"improving":[121],"model":[122],"performance.Crucially,":[123],"further":[124],"confirms":[126],"effectiveness":[129],"of":[130,135],"this":[131],"independent":[134],"its":[136,145],"position":[137],"characterized":[139],"by":[140],"negligible":[141],"semantic":[142],"content,":[143],"validating":[144],"role":[146],"robust":[149],"dedicated":[151],"structural":[152],"sink.Our":[153],"code":[154],"available":[156],"at":[157],"https://github.com/":[158],"skywalker0523/OneTokenIsEnough.":[159]},"counts_by_year":[],"updated_date":"2026-08-26T07:47:46.906454","created_date":"2026-07-02T00:00:00"}