{"id":"https://openalex.org/W3034025298","doi":"https://doi.org/10.1145/3397271.3401195","title":"A Pairwise Probe for Understanding BERT Fine-Tuning on Machine Reading Comprehension","display_name":"A Pairwise Probe for Understanding BERT Fine-Tuning on Machine Reading Comprehension","publication_year":2020,"publication_date":"2020-07-25","ids":{"openalex":"https://openalex.org/W3034025298","doi":"https://doi.org/10.1145/3397271.3401195","mag":"3034025298"},"language":"en","primary_location":{"id":"doi:10.1145/3397271.3401195","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3397271.3401195","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval","raw_type":"proceedings-article"},"type":"article","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2006.01346","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Jie Cai","orcid":null},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Jie Cai","raw_affiliation_strings":["Peking University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Peking University, Beijing, China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Zhengzhou Zhu","orcid":null},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhengzhou Zhu","raw_affiliation_strings":["Peking University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Peking University, Beijing, China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Ping Nie","orcid":null},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ping Nie","raw_affiliation_strings":["Peking University, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Peking University, Beijing, China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"last","author":{"id":null,"display_name":"Qian Liu","orcid":null},"institutions":[{"id":"https://openalex.org/I114017466","display_name":"University of Technology Sydney","ror":"https://ror.org/03f0f6041","country_code":"AU","type":"education","lineage":["https://openalex.org/I114017466"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Qian Liu","raw_affiliation_strings":["University of Technology Sydney, Sydney, Australia"],"affiliations":[{"raw_affiliation_string":"University of Technology Sydney, Sydney, Australia","institution_ids":["https://openalex.org/I114017466"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I20231570"],"apc_list":null,"apc_paid":null,"fwci":0.274,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.63249444,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":96},"biblio":{"volume":null,"issue":null,"first_page":"1665","last_page":"1668"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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":0.9998999834060669,"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":0.9998999834060669,"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/T13629","display_name":"Text Readability and Simplification","score":0.9952999949455261,"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/coreference","display_name":"Coreference","score":0.9366999864578247},{"id":"https://openalex.org/keywords/pairwise-comparison","display_name":"Pairwise comparison","score":0.7800999879837036},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.5719000101089478},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.4916999936103821},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.4708000123500824},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.45410001277923584},{"id":"https://openalex.org/keywords/comprehension","display_name":"Comprehension","score":0.43860000371932983},{"id":"https://openalex.org/keywords/textual-entailment","display_name":"Textual entailment","score":0.35350000858306885}],"concepts":[{"id":"https://openalex.org/C28076734","wikidata":"https://www.wikidata.org/wiki/Q63087","display_name":"Coreference","level":3,"score":0.9366999864578247},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7900999784469604},{"id":"https://openalex.org/C184898388","wikidata":"https://www.wikidata.org/wiki/Q1435712","display_name":"Pairwise comparison","level":2,"score":0.7800999879837036},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.734000027179718},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.7121000289916992},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5719000101089478},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.4916999936103821},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.4708000123500824},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.45410001277923584},{"id":"https://openalex.org/C511192102","wikidata":"https://www.wikidata.org/wiki/Q5156948","display_name":"Comprehension","level":2,"score":0.43860000371932983},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.38760000467300415},{"id":"https://openalex.org/C95318506","wikidata":"https://www.wikidata.org/wiki/Q6588467","display_name":"Textual entailment","level":3,"score":0.35350000858306885},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.3483000099658966},{"id":"https://openalex.org/C554936623","wikidata":"https://www.wikidata.org/wiki/Q199657","display_name":"Reading (process)","level":2,"score":0.3456999957561493},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.32919999957084656},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.3077999949455261},{"id":"https://openalex.org/C186644900","wikidata":"https://www.wikidata.org/wiki/Q194152","display_name":"Parsing","level":2,"score":0.30329999327659607},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.298799991607666},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.2809000015258789},{"id":"https://openalex.org/C198942812","wikidata":"https://www.wikidata.org/wiki/Q496618","display_name":"Semantic property","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C203005215","wikidata":"https://www.wikidata.org/wiki/Q79798","display_name":"Machine translation","level":2,"score":0.27709999680519104},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2703999876976013},{"id":"https://openalex.org/C130318100","wikidata":"https://www.wikidata.org/wiki/Q2268914","display_name":"Semantic similarity","level":2,"score":0.25450000166893005}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1145/3397271.3401195","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3397271.3401195","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2006.01346","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2006.01346","pdf_url":"https://arxiv.org/pdf/2006.01346","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2006.01346","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2006.01346","pdf_url":"https://arxiv.org/pdf/2006.01346","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":16,"referenced_works":["https://openalex.org/W2896457183","https://openalex.org/W2908854766","https://openalex.org/W2912924812","https://openalex.org/W2946417913","https://openalex.org/W2946434750","https://openalex.org/W2948947170","https://openalex.org/W2949493644","https://openalex.org/W2950813464","https://openalex.org/W2963903928","https://openalex.org/W2964032269","https://openalex.org/W2964204621","https://openalex.org/W2972324944","https://openalex.org/W2975059944","https://openalex.org/W2989276524","https://openalex.org/W2989464093","https://openalex.org/W4288086191"],"related_works":[],"abstract_inverted_index":{"Pre-trained":[0],"models":[1],"have":[2,11],"brought":[3],"significant":[4],"improvements":[5],"to":[6,45,68,119,138,180],"many":[7],"NLP":[8],"tasks":[9,98],"and":[10,48,73,79,111,153,169,173,192,195,214],"been":[12],"extensively":[13],"analyzed.":[14],"But":[15],"little":[16,187],"is":[17,75,209],"known":[18],"about":[19],"the":[20,93,124,144,161,190,220],"effect":[21,188],"of":[22,39,103,146,151,163],"fine-tuning":[23,49,122],"on":[24,85,123,189],"specific":[25,201],"tasks.":[26,198],"Intuitively,":[27],"people":[28],"may":[29],"agree":[30],"that":[31,95],"a":[32,76,112,116,171],"pre-trained":[33,152,212],"model":[34,167],"already":[35],"learns":[36],"semantic":[37,197],"representations":[38],"words":[40],"(e.g.":[41,59,105],"synonyms":[42],"are":[43,81,217],"closer":[44],"each":[46,147],"other)":[47],"further":[50],"improves":[51],"its":[52],"capabilities":[53],"which":[54],"require":[55],"more":[56],"complicated":[57],"reasoning":[58],"coreference":[60,106],"resolution,":[61],"entity":[62,110],"boundary":[63],"detection,":[64],"etc).":[65],"However,":[66],"how":[67],"verify":[69],"these":[70],"arguments":[71],"analytically":[72],"quantitatively":[74],"challenging":[77],"task":[78],"there":[80],"few":[82],"works":[83],"focus":[84],"this":[86,89],"topic.":[87],"In":[88],"paper,":[90],"inspired":[91],"by":[92],"observation":[94],"most":[96],"probing":[97,140],"involve":[99],"identifying":[100],"matched":[101],"pairs":[102],"phrases":[104],"requires":[107],"matching":[108],"an":[109],"pronoun),":[113],"we":[114,131,142],"propose":[115],"pairwise":[117,139,158],"probe":[118,159],"understand":[120],"BERT":[121,208,213],"machine":[125],"reading":[126],"comprehension":[127],"(MRC)":[128],"task.":[129],"Specifically,":[130],"identify":[132],"five":[133],"phenomena":[134],"in":[135],"MRC.":[136],"According":[137],"tasks,":[141,206],"compare":[143],"performance":[145],"layer's":[148],"hidden":[149],"representation":[150],"fine-tuned":[154,207],"BERT.":[155],"The":[156],"proposed":[157],"alleviates":[160],"problem":[162],"distraction":[164],"from":[165],"inaccurate":[166],"training":[168],"makes":[170],"robust":[172],"quantitative":[174],"comparison.":[175],"Our":[176],"experimental":[177],"analysis":[178],"leads":[179],"highly":[181],"confident":[182],"conclusions:":[183],"(1)":[184],"Fine-tuning":[185],"has":[186],"fundamental":[191],"low-level":[193],"information":[194],"general":[196],"(2)":[199],"For":[200],"abilities":[202],"required":[203],"for":[204],"downstream":[205],"better":[210],"than":[211],"such":[215],"gaps":[216],"obvious":[218],"after":[219],"fifth":[221],"layer.":[222]},"counts_by_year":[{"year":2022,"cited_by_count":2}],"updated_date":"2026-03-07T16:01:11.037858","created_date":"2020-06-12T00:00:00"}