{"id":"https://openalex.org/W4290994975","doi":"https://doi.org/10.1162/tacl_a_00493","title":"Saturated Transformers are Constant-Depth Threshold Circuits","display_name":"Saturated Transformers are Constant-Depth Threshold Circuits","publication_year":2022,"publication_date":"2022-01-01","ids":{"openalex":"https://openalex.org/W4290994975","doi":"https://doi.org/10.1162/tacl_a_00493"},"language":"en","primary_location":{"id":"doi:10.1162/tacl_a_00493","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00493","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00493/2038506/tacl_a_00493.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00493/2038506/tacl_a_00493.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5109946254","display_name":"William Merrill","orcid":null},"institutions":[{"id":"https://openalex.org/I4210140341","display_name":"Allen Institute","ror":"https://ror.org/03cpe7c52","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I4210140341"]},{"id":"https://openalex.org/I57206974","display_name":"New York University","ror":"https://ror.org/0190ak572","country_code":"US","type":"education","lineage":["https://openalex.org/I57206974"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"William Merrill","raw_affiliation_strings":["Allen Institute for AI, USA. willm@nyu.edu","New York University, USA"],"affiliations":[{"raw_affiliation_string":"Allen Institute for AI, USA. willm@nyu.edu","institution_ids":["https://openalex.org/I4210140341"]},{"raw_affiliation_string":"New York University, USA","institution_ids":["https://openalex.org/I57206974"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077726785","display_name":"Ashish Sabharwal","orcid":null},"institutions":[{"id":"https://openalex.org/I4210140341","display_name":"Allen Institute","ror":"https://ror.org/03cpe7c52","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I4210140341"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Ashish Sabharwal","raw_affiliation_strings":["Allen Institute for AI, USA. ashishs@allenai.org"],"affiliations":[{"raw_affiliation_string":"Allen Institute for AI, USA. ashishs@allenai.org","institution_ids":["https://openalex.org/I4210140341"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5088517824","display_name":"Noah A. Smith","orcid":"https://orcid.org/0000-0002-2310-6380"},"institutions":[{"id":"https://openalex.org/I201448701","display_name":"University of Washington","ror":"https://ror.org/00cvxb145","country_code":"US","type":"education","lineage":["https://openalex.org/I201448701"]},{"id":"https://openalex.org/I4210140341","display_name":"Allen Institute","ror":"https://ror.org/03cpe7c52","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I4210140341"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Noah A. Smith","raw_affiliation_strings":["Allen Institute for AI, USA. noah@allenai.org","University of Washington, USA"],"affiliations":[{"raw_affiliation_string":"Allen Institute for AI, USA. noah@allenai.org","institution_ids":["https://openalex.org/I4210140341"]},{"raw_affiliation_string":"University of Washington, USA","institution_ids":["https://openalex.org/I201448701"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5077726785","https://openalex.org/A5088517824","https://openalex.org/A5109946254"],"corresponding_institution_ids":["https://openalex.org/I201448701","https://openalex.org/I4210140341","https://openalex.org/I57206974"],"apc_list":null,"apc_paid":null,"fwci":2.3449,"has_fulltext":true,"cited_by_count":18,"citation_normalized_percentile":{"value":0.9004557,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":89,"max":99},"biblio":{"volume":"10","issue":null,"first_page":"843","last_page":"856"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9995999932289124,"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":0.9995999932289124,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9977999925613403,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.9972000122070312,"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/transformer","display_name":"Transformer","score":0.7976453304290771},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7048431038856506},{"id":"https://openalex.org/keywords/electronic-circuit","display_name":"Electronic circuit","score":0.5327767133712769},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.33000248670578003},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.32476407289505005},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.3153771758079529},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.09074446558952332}],"concepts":[{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.7976453304290771},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7048431038856506},{"id":"https://openalex.org/C134146338","wikidata":"https://www.wikidata.org/wiki/Q1815901","display_name":"Electronic circuit","level":2,"score":0.5327767133712769},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.33000248670578003},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.32476407289505005},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.3153771758079529},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.09074446558952332},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1162/tacl_a_00493","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00493","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00493/2038506/tacl_a_00493.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},{"id":"pmh:oai:doaj.org/article:1ebb220d5f95428f9c13e8ad5c4304f4","is_oa":true,"landing_page_url":"https://doaj.org/article/1ebb220d5f95428f9c13e8ad5c4304f4","pdf_url":null,"source":{"id":"https://openalex.org/S112646816","display_name":"SHILAP Revista de lepidopterolog\u00eda","issn_l":"0300-5267","issn":["0300-5267","2340-4078"],"is_oa":true,"is_in_doaj":true,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Transactions of the Association for Computational Linguistics, Vol 10 (2022)","raw_type":"article"}],"best_oa_location":{"id":"doi:10.1162/tacl_a_00493","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00493","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00493/2038506/tacl_a_00493.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","score":0.7900000214576721,"display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4290994975.pdf","grobid_xml":"https://content.openalex.org/works/W4290994975.grobid-xml"},"referenced_works_count":31,"referenced_works":["https://openalex.org/W134510267","https://openalex.org/W1506378440","https://openalex.org/W2014850486","https://openalex.org/W2401138554","https://openalex.org/W2612690371","https://openalex.org/W2889175957","https://openalex.org/W2908802752","https://openalex.org/W2952744660","https://openalex.org/W2963059228","https://openalex.org/W2972515356","https://openalex.org/W3014096773","https://openalex.org/W3098666169","https://openalex.org/W3104298168","https://openalex.org/W3121462755","https://openalex.org/W3133204183","https://openalex.org/W3214210379","https://openalex.org/W4214566146","https://openalex.org/W4224874866","https://openalex.org/W4234348737","https://openalex.org/W4238846128","https://openalex.org/W4239263507","https://openalex.org/W4298227433","https://openalex.org/W6677900116","https://openalex.org/W6713024538","https://openalex.org/W6739901393","https://openalex.org/W6755207826","https://openalex.org/W6771550025","https://openalex.org/W6780226713","https://openalex.org/W6788417511","https://openalex.org/W6791675639","https://openalex.org/W6795898621"],"related_works":["https://openalex.org/W2748952813","https://openalex.org/W2051487156","https://openalex.org/W2073681303","https://openalex.org/W2544423928","https://openalex.org/W2053286651","https://openalex.org/W2181743346","https://openalex.org/W2187401768","https://openalex.org/W2181413294","https://openalex.org/W2052122378","https://openalex.org/W2864363823"],"abstract_inverted_index":{"Abstract":[0],"Transformers":[1],"have":[2],"become":[3],"a":[4,57,83],"standard":[5],"neural":[6],"network":[7],"architecture":[8],"for":[9],"many":[10],"NLP":[11],"problems,":[12],"motivating":[13],"theoretical":[14],"analysis":[15],"of":[16,21,65,78,85,109],"their":[17],"power":[18,37],"in":[19,36,68,96],"terms":[20],"formal":[22,137],"languages.":[23],"Recent":[24],"work":[25],"has":[26],"shown":[27],"that":[28,88,102],"transformers":[29,79,104,116],"with":[30,80,117],"hard":[31,54,86],"attention":[32,55,87,93],"are":[33],"quite":[34],"limited":[35],"(Hahn,":[38],"2020),":[39],"as":[40,131],"they":[41,139],"can":[42,120],"be":[43,121],"simulated":[44,122],"by":[45,123],"constant-depth":[46,124],"AND/OR":[47],"circuits":[48],"(Hao":[49],"et":[50],"al.,":[51],"2022).":[52],"However,":[53],"is":[56],"strong":[58],"assumption,":[59],"which":[60],"may":[61],"complicate":[62],"the":[63,75,92,106,128,136],"relevance":[64],"these":[66],"results":[67],"practice.":[69],"In":[70],"this":[71],"work,":[72],"we":[73],"analyze":[74],"circuit":[76],"complexity":[77],"saturated":[81,103,115],"attention:":[82],"generalization":[84],"more":[89],"closely":[90],"captures":[91],"patterns":[94],"learnable":[95],"practical":[97],"transformers.":[98,111],"We":[99,112],"first":[100],"show":[101],"transcend":[105],"known":[107],"limitations":[108],"hard-attention":[110],"then":[113],"prove":[114],"floating-point":[118],"values":[119],"threshold":[125],"circuits,":[126],"giving":[127],"class":[129],"TC0":[130],"an":[132],"upper":[133],"bound":[134],"on":[135],"languages":[138],"recognize.":[140]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":1},{"year":2023,"cited_by_count":11},{"year":2022,"cited_by_count":1}],"updated_date":"2026-03-27T14:29:43.386196","created_date":"2025-10-10T00:00:00"}