iBet uBet
web content aggregator. Adding the entire web to your favor.
Link to original content:
https://api.openalex.org/works/doi:10.48550/ARXIV.2403.09963
{"id":"https://openalex.org/W4392929841","doi":"https://doi.org/10.48550/arxiv.2403.09963","title":"Take Care of Your Prompt Bias! Investigating and Mitigating Prompt Bias\n in Factual Knowledge Extraction","display_name":"Take Care of Your Prompt Bias! Investigating and Mitigating Prompt Bias\n in Factual Knowledge Extraction","publication_year":2024,"publication_date":"2024-03-14","ids":{"openalex":"https://openalex.org/W4392929841","doi":"https://doi.org/10.48550/arxiv.2403.09963"},"language":"en","primary_location":{"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.09963","pdf_url":"http://arxiv.org/pdf/2403.09963","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":["Cornell University"],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false},"type":"preprint","type_crossref":"posted-content","indexed_in":["arxiv"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"http://arxiv.org/pdf/2403.09963","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101447684","display_name":"Ziyang Xu","orcid":"https://orcid.org/0000-0001-7293-2999"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Ziyang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5094185912","display_name":"Keqin Peng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Peng, Keqin","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113438097","display_name":"Liang Ding","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ding, Liang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5094185913","display_name":"Dacheng Tao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tao, Dacheng","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5111234377","display_name":"Xiliang Lu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu, Xiliang","raw_affiliation_strings":[],"affiliations":[]}],"institution_assertions":[],"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":1,"citation_normalized_percentile":{"value":0.999951,"is_in_top_1_percent":true,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":85,"max":94},"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Natural Language Processing","score":0.9873,"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":"Natural Language Processing","score":0.9873,"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":"Statistical Machine Translation and Natural Language Processing","score":0.9627,"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/T10260","display_name":"Empirical Studies in Software Engineering","score":0.9457,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/topic-modeling","display_name":"Topic Modeling","score":0.533187},{"id":"https://openalex.org/keywords/language-modeling","display_name":"Language Modeling","score":0.533133},{"id":"https://openalex.org/keywords/pretrained-models","display_name":"Pretrained Models","score":0.515649},{"id":"https://openalex.org/keywords/software-defect-prediction","display_name":"Software Defect Prediction","score":0.5084}],"concepts":[{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.4685394},{"id":"https://openalex.org/C4725764","wikidata":"https://www.wikidata.org/wiki/Q844704","display_name":"Extraction (chemistry)","level":2,"score":0.44249463},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.33109653},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.32886362},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.11183414},{"id":"https://openalex.org/C43617362","wikidata":"https://www.wikidata.org/wiki/Q170050","display_name":"Chromatography","level":1,"score":0.09475684}],"mesh":[],"locations_count":1,"locations":[{"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.09963","pdf_url":"http://arxiv.org/pdf/2403.09963","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":["Cornell University"],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false}],"best_oa_location":{"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.09963","pdf_url":"http://arxiv.org/pdf/2403.09963","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":["Cornell University"],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false},"sustainable_development_goals":[],"grants":[],"datasets":[],"versions":[],"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391913857","https://openalex.org/W2748952813","https://openalex.org/W2530322880","https://openalex.org/W2478288626","https://openalex.org/W2390279801","https://openalex.org/W2382290278","https://openalex.org/W2376932109","https://openalex.org/W2358668433","https://openalex.org/W2350741829","https://openalex.org/W2001405890"],"abstract_inverted_index":{"Recent":[0],"research":[1],"shows":[2],"that":[3,159],"pre-trained":[4],"language":[5],"models":[6],"(PLMs)":[7],"suffer":[8],"from":[9,132],"\"prompt":[10],"bias\"":[11],"in":[12,63,200,223],"factual":[13],"knowledge":[14,216],"extraction,":[15],"i.e.,":[16],"prompts":[17,49,62,71],"tend":[18],"to":[19,110,137,145,182,211],"introduce":[20],"biases":[21],"toward":[22,214],"specific":[23],"labels.":[24],"However,":[25],"the":[26,34,43,64,92,112,122,133,139,147,166,177,193],"extent":[27],"and":[28,50,74,128,156,219],"impact":[29,53],"of":[30,48,80,197],"prompt":[31,83,113,171,178],"bias":[32,44,84,114],"within":[33],"model":[35],"remain":[36],"underexplored.":[37],"In":[38],"response,":[39],"this":[40],"paper":[41],"quantifies":[42],"with":[45,69],"various":[46,153],"types":[47],"assesses":[51],"their":[52],"on":[54,96,102,192],"different":[55],"benchmarks.":[56],"We":[57],"show":[58,158],"that:":[59],"1)":[60],"all":[61],"experiments":[65],"exhibit":[66],"non-negligible":[67],"bias,":[68,172],"gradient-based":[70],"like":[72,99],"AutoPrompt":[73],"OptiPrompt":[75],"displaying":[76],"significantly":[77,175],"higher":[78],"levels":[79],"bias;":[81],"2)":[82],"can":[85,162,206],"amplify":[86],"benchmark":[87],"accuracy":[88],"unreasonably":[89],"by":[90,170],"overfitting":[91],"test":[93],"datasets,":[94],"especially":[95],"imbalanced":[97],"datasets":[98],"LAMA.":[100],"Based":[101],"these":[103],"findings,":[104],"we":[105,119],"propose":[106],"a":[107,208],"representation-based":[108],"approach":[109,161,205],"mitigate":[111],"during":[115],"inference":[116],"time.":[117],"Specifically,":[118],"first":[120],"estimate":[121],"biased":[123],"representation":[124],"using":[125],"prompt-only":[126],"querying,":[127],"then":[129],"remove":[130],"it":[131],"model's":[134],"internal":[135],"representations":[136],"generate":[138],"debiased":[140,149],"representations,":[141],"which":[142],"are":[143,221],"used":[144],"produce":[146],"final":[148],"outputs.":[150],"Experiments":[151],"across":[152],"prompts,":[154],"PLMs,":[155],"benchmarks":[157],"our":[160,203],"not":[163],"only":[164],"correct":[165],"overfitted":[167],"performance":[168,185],"caused":[169],"but":[173],"also":[174],"improve":[176],"retrieval":[179],"capability":[180],"(up":[181],"10%":[183],"absolute":[184],"gain).":[186],"Our":[187],"findings":[188],"shed":[189],"new":[190],"light":[191],"underlying":[194],"predicting":[195],"mechanisms":[196],"prompt-based":[198],"queries":[199],"PLMs.":[201],"Hopefully,":[202],"plug-and-play":[204],"be":[207],"golden":[209],"standard":[210],"strengthen":[212],"PLMs":[213],"reliable":[215],"bases.":[217],"Code":[218],"data":[220],"released":[222],"https://github.com/FelliYang/PromptBias.":[224]},"cited_by_api_url":"https://api.openalex.org/works?filter=cites:W4392929841","counts_by_year":[{"year":2024,"cited_by_count":1}],"updated_date":"2024-11-26T12:28:31.183517","created_date":"2024-03-19"}
