{"id":"https://openalex.org/W7126425383","doi":"https://doi.org/10.18653/v1/2024.findings-eacl.91","title":"Prompt Perturbation Consistency Learning for Robust Language Models","display_name":"Prompt Perturbation Consistency Learning for Robust Language Models","publication_year":2024,"publication_date":"2024-01-01","ids":{"openalex":"https://openalex.org/W7126425383","doi":"https://doi.org/10.18653/v1/2024.findings-eacl.91"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2024.findings-eacl.91","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.findings-eacl.91","pdf_url":"https://aclanthology.org/2024.findings-eacl.91.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":"Findings of the Association for Computational Linguistics: EACL 2024","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2024.findings-eacl.91.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5018415209","display_name":"Yao Qiang","orcid":"https://orcid.org/0000-0003-2995-3385"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yao Qiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5086592108","display_name":"Subhrangshu Nandi","orcid":"https://orcid.org/0009-0008-4855-2548"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Subhrangshu Nandi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056269049","display_name":"Ninareh Mehrabi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ninareh Mehrabi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124633415","display_name":"Greg Ver Steeg","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Greg Ver Steeg","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124690801","display_name":"Anoop Kumar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Anoop Kumar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071360545","display_name":"Anna Rumshisky","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Anna Rumshisky","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5101715504","display_name":"Aram Galstyan","orcid":"https://orcid.org/0000-0003-4215-0886"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Aram Galstyan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":7,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.57284451,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1357","last_page":"1370"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.3978999853134155,"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":0.3978999853134155,"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":0.14100000262260437,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.07739999890327454,"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/consistency","display_name":"Consistency (knowledge bases)","score":0.3749000132083893},{"id":"https://openalex.org/keywords/perturbation","display_name":"Perturbation (astronomy)","score":0.3621000051498413},{"id":"https://openalex.org/keywords/strong-consistency","display_name":"Strong consistency","score":0.29510000348091125},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.2831000089645386}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5317999720573425},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.42489999532699585},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.3749000132083893},{"id":"https://openalex.org/C177918212","wikidata":"https://www.wikidata.org/wiki/Q803623","display_name":"Perturbation (astronomy)","level":2,"score":0.3621000051498413},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3393000066280365},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.32510000467300415},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3021000027656555},{"id":"https://openalex.org/C31388003","wikidata":"https://www.wikidata.org/wiki/Q7624548","display_name":"Strong consistency","level":3,"score":0.29510000348091125},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2831000089645386},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.25189998745918274}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2024.findings-eacl.91","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.findings-eacl.91","pdf_url":"https://aclanthology.org/2024.findings-eacl.91.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":"Findings of the Association for Computational Linguistics: EACL 2024","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2024.findings-eacl.91","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.findings-eacl.91","pdf_url":"https://aclanthology.org/2024.findings-eacl.91.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":"Findings of the Association for Computational Linguistics: EACL 2024","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7126425383.pdf","grobid_xml":"https://content.openalex.org/works/W7126425383.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1,13],"models":[2,99],"(LLMs)":[3],"have":[4],"demonstrated":[5],"impressive":[6],"performance":[7,24,85,94,152],"on":[8,25,58,144],"a":[9,39,53],"number":[10],"of":[11,55,60,70,96,107,150],"natural":[12],"processing":[14],"tasks,":[15,28,158],"such":[16,29],"as":[17,30],"question":[18],"answering":[19],"and":[20,33,112,135,148,156],"text":[21],"summarization.However,":[22],"their":[23],"sequence":[26],"labeling":[27],"intent":[31],"classification":[32],"slot":[34],"filling":[35],"(IC-SF),":[36],"which":[37,125],"is":[38,52],"central":[40],"component":[41],"in":[42,65],"personal":[43],"assistant":[44],"systems,":[45],"lags":[46],"significantly":[47],"behind":[48],"discriminative":[49,88],"models.Furthermore,":[50],"there":[51],"lack":[54],"substantive":[56],"research":[57],"robustness":[59],"LLMs":[61,81],"to":[62,87,101],"various":[63],"perturbations":[64,109],"the":[66,93,129,151],"input":[67,108],"prompts.The":[68],"contributions":[69],"this":[71],"paper":[72],"are":[73],"three-fold.First,":[74],"we":[75,90,114],"show":[76,139],"that":[77,140],"finetuning":[78],"sufficiently":[79],"large":[80],"can":[82,142],"produce":[83],"IC-SF":[84],"comparable":[86],"models.Next,":[89],"systematically":[91],"analyze":[92],"deterioration":[95],"those":[97],"fine-tuned":[98],"due":[100],"three":[102],"distinct":[103],"yet":[104],"relevant":[105],"types":[106],"-oronyms,":[110],"synonyms,":[111],"paraphrasing.Finally,":[113],"propose":[115],"an":[116,145],"efficient":[117],"mitigation":[118],"approach,":[119],"prompt":[120],"perturbation":[121],"consistency":[122],"learning":[123],"(PPCL),":[124],"works":[126],"by":[127],"regularizing":[128],"divergence":[130],"between":[131],"losses":[132],"from":[133],"clean":[134],"perturbed":[136],"samples.Our":[137],"experiments":[138],"PPCL":[141,160],"recover":[143],"average":[146],"59%":[147],"69%":[149],"drop":[153],"for":[154],"IC":[155],"SF":[157],"respectively.Furthermore,":[159],"beats":[161],"data":[162,171],"augmentation":[163],"approach":[164],"while":[165],"using":[166],"ten":[167],"times":[168],"fewer":[169],"augmented":[170],"samples.":[172]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-02-02T00:00:00"}