{"id":"https://openalex.org/W3133029875","doi":"https://doi.org/10.48550/arxiv.2102.13019","title":"Investigating the Limitations of Transformers with Simple Arithmetic Tasks","display_name":"Investigating the Limitations of Transformers with Simple Arithmetic Tasks","publication_year":2021,"publication_date":"2021-02-25","ids":{"openalex":"https://openalex.org/W3133029875","doi":"https://doi.org/10.48550/arxiv.2102.13019","mag":"3133029875"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2102.13019","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2102.13019","pdf_url":"https://arxiv.org/pdf/2102.13019","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2102.13019","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5030647281","display_name":"Rodrigo Frassetto Nogueira","orcid":"https://orcid.org/0000-0002-2600-6035","observed_orcids":["https://orcid.org/0000-0002-2600-6035"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nogueira, Rodrigo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101513734","display_name":"Zhiying Jiang","orcid":"https://orcid.org/0000-0001-8840-7056","observed_orcids":["https://orcid.org/0000-0001-8840-7056"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Zhiying","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5082997975","display_name":"Jimmy Lin","orcid":"https://orcid.org/0000-0002-0661-7189","observed_orcids":["https://orcid.org/0000-0002-0661-7189","https://orcid.org/0009-0007-1571-3854"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Jimmy","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":false,"cited_by_count":29,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.8384000062942505,"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":0.8384000062942505,"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":0.13179999589920044,"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":0.0032999999821186066,"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/arithmetic-reasoning","display_name":"arithmetic reasoning","score":0.7170000076293945},{"id":"https://openalex.org/keywords/addition","display_name":"addition","score":0.503000020980835},{"id":"https://openalex.org/keywords/subtraction","display_name":"subtraction","score":0.9990000128746033},{"id":"https://openalex.org/keywords/sequence-to-sequence-models","display_name":"sequence-to-sequence models","score":0.7459999918937683},{"id":"https://openalex.org/keywords/subword-tokenization","display_name":"subword tokenization","score":0.9229999780654907},{"id":"https://openalex.org/keywords/character-level-tokenization","display_name":"character-level tokenization","score":0.703000009059906},{"id":"https://openalex.org/keywords/positional-encoding","display_name":"positional encoding","score":0.8569999933242798},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.6290000081062317}],"concepts":[{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.7356722354888916},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5722526907920837},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.5589940547943115},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4636012613773346},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.33481794595718384},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.1439356505870819},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.1283450424671173},{"id":"https://openalex.org/C111472728","wikidata":"https://www.wikidata.org/wiki/Q9471","display_name":"Epistemology","level":1,"score":0.10433852672576904},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.08234763145446777},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0755394697189331}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2102.13019","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2102.13019","pdf_url":"https://arxiv.org/pdf/2102.13019","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2102.13019","doi":"https://doi.org/10.48550/arxiv.2102.13019","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2102.13019","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2102.13019","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2102.13019","pdf_url":"https://arxiv.org/pdf/2102.13019","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":47,"referenced_works":["https://openalex.org/W1732222442","https://openalex.org/W1771459135","https://openalex.org/W2173051530","https://openalex.org/W2548137223","https://openalex.org/W2626778328","https://openalex.org/W2766736793","https://openalex.org/W2787560479","https://openalex.org/W2866343820","https://openalex.org/W2887020936","https://openalex.org/W2896457183","https://openalex.org/W2908510526","https://openalex.org/W2910937705","https://openalex.org/W2919420119","https://openalex.org/W2946417913","https://openalex.org/W2950645060","https://openalex.org/W2951107864","https://openalex.org/W2951756287","https://openalex.org/W2963005248","https://openalex.org/W2970308008","https://openalex.org/W2970609357","https://openalex.org/W2970900584","https://openalex.org/W2971094176","https://openalex.org/W2974596953","https://openalex.org/W2981037730","https://openalex.org/W2994278562","https://openalex.org/W2995971510","https://openalex.org/W2998209000","https://openalex.org/W2998696444","https://openalex.org/W3015339775","https://openalex.org/W3022766797","https://openalex.org/W3024482470","https://openalex.org/W3033187248","https://openalex.org/W3034811314","https://openalex.org/W3082274269","https://openalex.org/W3083835029","https://openalex.org/W3092044512","https://openalex.org/W3092689172","https://openalex.org/W3093356033","https://openalex.org/W3095645723","https://openalex.org/W3098903812","https://openalex.org/W3100879603","https://openalex.org/W3102999298","https://openalex.org/W3106210592","https://openalex.org/W3111739346","https://openalex.org/W3122515622","https://openalex.org/W3128590981","https://openalex.org/W3138296101"],"related_works":["https://openalex.org/W4389115202","https://openalex.org/W7071739557","https://openalex.org/W7159554042","https://openalex.org/W4366835609","https://openalex.org/W4382491649","https://openalex.org/W7124319095","https://openalex.org/W4383860315","https://openalex.org/W7204166209","https://openalex.org/W4416036536","https://openalex.org/W4404088963"],"abstract_inverted_index":{"The":[0],"ability":[1],"to":[2,84,98,117,124,202],"perform":[3],"arithmetic":[4,46,137],"tasks":[5,47],"is":[6,65,206],"a":[7,16,34,54,63,72],"remarkable":[8],"trait":[9],"of":[10,19,33,57,87,176,179,192,195],"human":[11],"intelligence":[12],"and":[13,51,95,120,158,181],"might":[14,168],"form":[15,32,70],"critical":[17],"component":[18],"more":[20],"complex":[21],"reasoning":[22],"tasks.":[23],"In":[24,79],"this":[25],"work,":[26],"we":[27,145,172],"investigate":[28],"if":[29],"the":[30,76,81,114,147,177,193,196],"surface":[31,69,149],"number":[35,64,178],"has":[36,71],"any":[37],"influence":[38,74],"on":[39,75],"how":[40,62],"sequence-to-sequence":[41],"language":[42,132],"models":[43,133,184],"learn":[44,85,99,136,186],"simple":[45],"such":[48],"as":[49,142,144],"addition":[50,86,187],"subtraction":[52],"across":[53],"wide":[55],"range":[56],"values.":[58],"We":[59,127],"find":[60],"that":[61,129,155,167,174,189],"represented":[66],"in":[67,163],"its":[68],"strong":[73],"model's":[77],"accuracy.":[78],"particular,":[80],"model":[82,115],"fails":[83],"five-digit":[88],"numbers":[89,122,197],"when":[90],"using":[91],"subwords":[92],"(e.g.,":[93,103,110],"\"32\"),":[94],"it":[96],"struggles":[97],"with":[100],"character-level":[101],"representations":[102],"\"3":[104,111],"2\").":[105],"By":[106],"introducing":[107],"position":[108],"tokens":[109],"10e1":[112],"2\"),":[113],"learns":[116],"accurately":[118],"add":[119],"subtract":[121],"up":[123],"60":[125],"digits.":[126],"conclude":[128],"modern":[130],"pretrained":[131],"can":[134],"easily":[135],"from":[138],"very":[139],"few":[140],"examples,":[141,183],"long":[143],"use":[146],"proper":[148],"representation.":[150],"This":[151],"result":[152],"bolsters":[153],"evidence":[154],"subword":[156],"tokenizers":[157],"positional":[159],"encodings":[160],"are":[161,190],"components":[162],"current":[164],"transformer":[165],"designs":[166],"need":[169],"improvement.":[170],"Moreover,":[171],"show":[173],"regardless":[175],"parameters":[180],"training":[182],"cannot":[185],"rules":[188],"independent":[191],"length":[194],"seen":[198],"during":[199],"training.":[200],"Code":[201],"reproduce":[203],"our":[204],"experiments":[205],"available":[207],"at":[208],"https://github.com/castorini/transformers-arithmetic":[209]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":2},{"year":2024,"cited_by_count":6},{"year":2023,"cited_by_count":5},{"year":2022,"cited_by_count":7},{"year":2021,"cited_by_count":8}],"updated_date":"2026-10-16T09:59:22.235173","created_date":"2021-03-01T00:00:00"}