{"id":"https://openalex.org/W4385573248","doi":"https://doi.org/10.18653/v1/2022.emnlp-main.41","title":"When Can Transformers Ground and Compose: Insights from Compositional Generalization Benchmarks","display_name":"When Can Transformers Ground and Compose: Insights from Compositional Generalization Benchmarks","publication_year":2022,"publication_date":"2022-01-01","ids":{"openalex":"https://openalex.org/W4385573248","doi":"https://doi.org/10.18653/v1/2022.emnlp-main.41"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2022.emnlp-main.41","is_oa":true,"landing_page_url":"http://dx.doi.org/10.18653/v1/2022.emnlp-main.41","pdf_url":"https://aclanthology.org/2022.emnlp-main.41.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 2022 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2022.emnlp-main.41.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5022074748","display_name":"Ankur Sikarwar","orcid":"https://orcid.org/0009-0007-4946-5289"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Ankur Sikarwar","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5015473787","display_name":"Arkil Patel","orcid":null},"institutions":[{"id":"https://openalex.org/I5023651","display_name":"McGill University","ror":"https://ror.org/01pxwe438","country_code":"CA","type":"education","lineage":["https://openalex.org/I5023651"]},{"id":"https://openalex.org/I4210164802","display_name":"Mila - Quebec Artificial Intelligence Institute","ror":"https://ror.org/05c22rx21","country_code":"CA","type":"facility","lineage":["https://openalex.org/I4210164802"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Arkil Patel","raw_affiliation_strings":["McGill University","Mila -Quebec AI Institute"],"affiliations":[{"raw_affiliation_string":"McGill University","institution_ids":["https://openalex.org/I5023651"]},{"raw_affiliation_string":"Mila -Quebec AI Institute","institution_ids":["https://openalex.org/I4210164802"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5029531416","display_name":"Navin Goyal","orcid":"https://orcid.org/0000-0002-8521-0108"},"institutions":[{"id":"https://openalex.org/I4210164802","display_name":"Mila - Quebec Artificial Intelligence Institute","ror":"https://ror.org/05c22rx21","country_code":"CA","type":"facility","lineage":["https://openalex.org/I4210164802"]},{"id":"https://openalex.org/I4210124949","display_name":"Microsoft Research (India)","ror":"https://ror.org/02w7f3w92","country_code":"IN","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210124949"]}],"countries":["CA","IN"],"is_corresponding":false,"raw_author_name":"Navin Goyal","raw_affiliation_strings":["Microsoft Research India","Mila -Quebec AI Institute"],"affiliations":[{"raw_affiliation_string":"Microsoft Research India","institution_ids":["https://openalex.org/I4210124949"]},{"raw_affiliation_string":"Mila -Quebec AI Institute","institution_ids":["https://openalex.org/I4210164802"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5022074748"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.5303,"has_fulltext":true,"cited_by_count":4,"citation_normalized_percentile":{"value":0.72276106,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"648","last_page":"669"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.991599977016449,"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":0.991599977016449,"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":0.9905999898910522,"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":0.9812999963760376,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.7833750247955322},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7379468083381653},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.6770754456520081},{"id":"https://openalex.org/keywords/grid","display_name":"Grid","score":0.5617910027503967},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.48086053133010864},{"id":"https://openalex.org/keywords/ground","display_name":"Ground","score":0.4600394368171692},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4297291934490204},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.3596777617931366},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.34870386123657227},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.2300836443901062},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.11412280797958374},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.106996089220047},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.10300076007843018}],"concepts":[{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.7833750247955322},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7379468083381653},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.6770754456520081},{"id":"https://openalex.org/C187691185","wikidata":"https://www.wikidata.org/wiki/Q2020720","display_name":"Grid","level":2,"score":0.5617910027503967},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.48086053133010864},{"id":"https://openalex.org/C168993435","wikidata":"https://www.wikidata.org/wiki/Q6501125","display_name":"Ground","level":2,"score":0.4600394368171692},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4297291934490204},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3596777617931366},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.34870386123657227},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2300836443901062},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.11412280797958374},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.106996089220047},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.10300076007843018},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.0},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2022.emnlp-main.41","is_oa":true,"landing_page_url":"http://dx.doi.org/10.18653/v1/2022.emnlp-main.41","pdf_url":"https://aclanthology.org/2022.emnlp-main.41.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 2022 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2022.emnlp-main.41","is_oa":true,"landing_page_url":"http://dx.doi.org/10.18653/v1/2022.emnlp-main.41","pdf_url":"https://aclanthology.org/2022.emnlp-main.41.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 2022 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[{"display_name":"Quality Education","score":0.7799999713897705,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[{"id":"https://openalex.org/F4320308943","display_name":"Microsoft Research","ror":"https://ror.org/00d0nc645"}],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4385573248.pdf"},"referenced_works_count":8,"referenced_works":["https://openalex.org/W2964118342","https://openalex.org/W3094502228","https://openalex.org/W3105725479","https://openalex.org/W3177813494","https://openalex.org/W3197009789","https://openalex.org/W3202070718","https://openalex.org/W4287995219","https://openalex.org/W4292779060"],"related_works":["https://openalex.org/W2021787609","https://openalex.org/W2097328689","https://openalex.org/W4234899305","https://openalex.org/W2054767487","https://openalex.org/W1537063595","https://openalex.org/W2379604501","https://openalex.org/W2373854414","https://openalex.org/W2954371137","https://openalex.org/W2371138613","https://openalex.org/W2392155297"],"abstract_inverted_index":{"Humans":[0],"can":[1,113],"reason":[2,133],"compositionally":[3],"whilst":[4],"grounding":[5],"language":[6],"utterances":[7],"to":[8,28,115,129,147],"the":[9,67,73,77,81,85,161,165,174,180],"real":[10],"world.":[11],"Recent":[12],"benchmarks":[13],"like":[14],"ReaSCAN":[15,50],"(Wu":[16],"et":[17,56,62],"al.,":[18,57,63],"2021)":[19,58],"use":[20],"navigation":[21],"tasks":[22],"grounded":[23,124,175],"in":[24,76,94,193],"a":[25,41,52,91,122,138,143,156],"grid":[26,78],"world":[27,79],"assess":[29],"whether":[30],"neural":[31],"models":[32],"exhibit":[33],"similar":[34],"capabilities.":[35],"In":[36],"this":[37,107,194],"work,":[38],"we":[39,69,88,109,120,154,169],"present":[40],"simple":[42],"transformer-based":[43],"model":[44],"that":[45,71,90,111,137],"outperforms":[46],"specialized":[47],"architectures":[48],"on":[49,184],"and":[51,179],"modified":[53],"version":[54,105],"(Qiu":[55],"of":[59,106,150,160,182],"gSCAN":[60],"(Ruis":[61],"2020).":[64],"On":[65,102],"analyzing":[66],"task,":[68,127],"find":[70],"identifying":[72],"target":[74],"location":[75],"is":[80,100],"main":[82],"challenge":[83],"for":[84,190],"models.":[86],"Furthermore,":[87],"show":[89,110,136],"particular":[92],"split":[93],"ReaSCAN,":[95],"which":[96,186],"tests":[97],"depth":[98],"generalization,":[99],"unfair.":[101],"an":[103],"amended":[104],"split,":[108],"transformers":[112,132,183],"generalize":[114],"deeper":[116],"input":[117],"structures.":[118],"Finally,":[119],"design":[121],"simpler":[123],"compositional":[125,176],"generalization":[126,177],"RefEx,":[128],"investigate":[130],"how":[131],"compositionally.":[134],"We":[135],"single":[139,144],"self-attention":[140],"layer":[141],"with":[142],"head":[145],"generalizes":[146],"novel":[148],"combinations":[149],"object":[151],"attributes.":[152],"Moreover,":[153],"derive":[155],"precise":[157],"mathematical":[158],"construction":[159],"transformer\u2019s":[162],"computations":[163],"from":[164],"learned":[166],"network.":[167],"Overall,":[168],"provide":[170],"valuable":[171],"insights":[172],"about":[173],"task":[178],"behaviour":[181],"it,":[185],"would":[187],"be":[188],"useful":[189],"researchers":[191],"working":[192],"area.":[195]},"counts_by_year":[{"year":2024,"cited_by_count":1},{"year":2023,"cited_by_count":3}],"updated_date":"2025-12-21T01:58:51.020947","created_date":"2025-10-10T00:00:00"}