{"id":"https://openalex.org/W4411630053","doi":"https://doi.org/10.18653/v1/2024.eacl-long.175","title":"Gradient-Based Language Model Red Teaming","display_name":"Gradient-Based Language Model Red Teaming","publication_year":2024,"publication_date":"2024-01-01","ids":{"openalex":"https://openalex.org/W4411630053","doi":"https://doi.org/10.18653/v1/2024.eacl-long.175"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2024.eacl-long.175","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.eacl-long.175","pdf_url":"https://aclanthology.org/2024.eacl-long.175.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 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2024.eacl-long.175.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5093836219","display_name":"Nevan Wichers","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Nevan Wichers","raw_affiliation_strings":["Google Research Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research Google Research","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079864698","display_name":"Carson Denison","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Carson Denison","raw_affiliation_strings":["Google Research Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research Google Research","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5008645615","display_name":"Ahmad Beirami","orcid":"https://orcid.org/0000-0002-1998-5271"},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ahmad Beirami","raw_affiliation_strings":["Google Research Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research Google Research","institution_ids":["https://openalex.org/I1291425158"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I1291425158"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":6,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"2862","last_page":"2881"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.8306999802589417,"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":0.8306999802589417,"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/T10215","display_name":"Semantic Web and Ontologies","score":0.8069000244140625,"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/T10203","display_name":"Recommender Systems and Techniques","score":0.7746000289916992,"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/computer-science","display_name":"Computer science","score":0.6589335799217224},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.4095768332481384},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.38682612776756287}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6589335799217224},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.4095768332481384},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.38682612776756287}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2024.eacl-long.175","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.eacl-long.175","pdf_url":"https://aclanthology.org/2024.eacl-long.175.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 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2024.eacl-long.175","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.eacl-long.175","pdf_url":"https://aclanthology.org/2024.eacl-long.175.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 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320323029","display_name":"Atomic Energy of Canada Limited","ror":"https://ror.org/019ja9e37"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4411630053.pdf","grobid_xml":"https://content.openalex.org/works/W4411630053.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Red":[0,52],"teaming":[1,27,57,159],"is":[2,28,37,75,139],"a":[3,55,76,88,117,122,154],"common":[4],"strategy":[5],"for":[6,30,59],"identifying":[7],"weaknesses":[8],"in":[9],"generative":[10],"language":[11],"models":[12],"(LMs),":[13],"where":[14],"adversarial":[15],"prompts":[16,63,128,133,144],"are":[17,65],"produced":[18],"that":[19,64,115,137,145],"trigger":[20,146],"an":[21,69,84,147],"LM":[22,70,85,100,148,166],"to":[23,41,67,71,101,125,149,170],"generate":[24,126,150],"unsafe":[25,73,151],"responses.Red":[26],"instrumental":[29],"both":[31],"model":[32,124],"alignment":[33],"and":[34,39,91,99,120,161],"evaluation,":[35],"but":[36],"laborintensive":[38],"difficult":[40],"scale":[42],"when":[43,164],"done":[44],"by":[45,82],"humans.In":[46],"this":[47],"paper,":[48],"we":[49,111],"present":[50],"Gradient-Based":[51],"Teaming":[53],"(GBRT),":[54],"red":[56,158],"method":[58],"automatically":[60],"generating":[61],"diverse":[62],"likely":[66],"cause":[68],"output":[72],"responses.GBRT":[74],"form":[77],"of":[78,108,130],"prompt":[79],"learning,":[80],"trained":[81],"scoring":[83],"response":[86],"with":[87],"safety":[89,97],"classifier":[90,98],"then":[92],"backpropagating":[93],"through":[94],"the":[95,103,106,127,132,165],"frozen":[96],"update":[102],"prompt.To":[104],"improve":[105],"coherence":[107],"input":[109],"prompts,":[110],"introduce":[112],"two":[113],"variants":[114],"add":[116],"realism":[118],"loss":[119],"fine-tune":[121],"pretrained":[123],"instead":[129],"learning":[131],"directly.Our":[134],"experiments":[135],"show":[136],"GBRT":[138],"more":[140],"effective":[141],"at":[142],"finding":[143],"responses":[152],"than":[153],"strong":[155],"reinforcement":[156],"learning-based":[157],"approach,":[160],"succeeds":[162],"even":[163],"has":[167],"been":[168],"fine-tuned":[169],"produce":[171],"safer":[172],"outputs.":[173],"1":[174]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":4}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}