{"id":"https://openalex.org/W4412888465","doi":"https://doi.org/10.18653/v1/2025.findings-acl.564","title":"Dynamic Evil Score-Guided Decoding: An Efficient Decoding Framework For Red-Team Model","display_name":"Dynamic Evil Score-Guided Decoding: An Efficient Decoding Framework For Red-Team Model","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888465","doi":"https://doi.org/10.18653/v1/2025.findings-acl.564"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.564","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.564","pdf_url":"https://aclanthology.org/2025.findings-acl.564.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: ACL 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.564.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100653319","display_name":"Cong Gao","orcid":"https://orcid.org/0000-0001-9829-8025"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cong Gao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100335250","display_name":"Bo Zhang","orcid":"https://orcid.org/0000-0002-1579-3707"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bo Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039888657","display_name":"Linkang Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Linkang Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101814775","display_name":"Minghao Hu","orcid":"https://orcid.org/0000-0002-8573-7110"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Minghao Hu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101037209","display_name":"Zhunchen Luo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhunchen Luo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062486535","display_name":"Xiaoying Bai","orcid":"https://orcid.org/0000-0003-3989-4075"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaoying Bai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070742322","display_name":"Guotong Geng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guotong Geng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100433085","display_name":"Jun Zhang","orcid":"https://orcid.org/0000-0001-9093-9430"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jun Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5081072099","display_name":"Yunhua Xue","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yunhua Xue","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"10817","last_page":"10833"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10734","display_name":"Information and Cyber Security","score":0.8095999956130981,"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"}},"topics":[{"id":"https://openalex.org/T10734","display_name":"Information and Cyber Security","score":0.8095999956130981,"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"}},{"id":"https://openalex.org/T12592","display_name":"Opinion Dynamics and Social Influence","score":0.8090000152587891,"subfield":{"id":"https://openalex.org/subfields/3109","display_name":"Statistical and Nonlinear Physics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10270","display_name":"Blockchain Technology Applications and Security","score":0.7919999957084656,"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/decoding-methods","display_name":"Decoding methods","score":0.8984404802322388},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6440027952194214},{"id":"https://openalex.org/keywords/sequential-decoding","display_name":"Sequential decoding","score":0.5646173357963562},{"id":"https://openalex.org/keywords/list-decoding","display_name":"List decoding","score":0.5630546808242798},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.36717987060546875},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.34556645154953003},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.2938534915447235},{"id":"https://openalex.org/keywords/concatenated-error-correction-code","display_name":"Concatenated error correction code","score":0.10877439379692078}],"concepts":[{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.8984404802322388},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6440027952194214},{"id":"https://openalex.org/C193969084","wikidata":"https://www.wikidata.org/wiki/Q7452500","display_name":"Sequential decoding","level":4,"score":0.5646173357963562},{"id":"https://openalex.org/C204397858","wikidata":"https://www.wikidata.org/wiki/Q4437907","display_name":"List decoding","level":5,"score":0.5630546808242798},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.36717987060546875},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.34556645154953003},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2938534915447235},{"id":"https://openalex.org/C78944582","wikidata":"https://www.wikidata.org/wiki/Q5158264","display_name":"Concatenated error correction code","level":4,"score":0.10877439379692078},{"id":"https://openalex.org/C157125643","wikidata":"https://www.wikidata.org/wiki/Q884707","display_name":"Block code","level":3,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.564","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.564","pdf_url":"https://aclanthology.org/2025.findings-acl.564.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: ACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.564","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.564","pdf_url":"https://aclanthology.org/2025.findings-acl.564.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: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888465.pdf","grobid_xml":"https://content.openalex.org/works/W4412888465.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4205451769","https://openalex.org/W3017203753","https://openalex.org/W2385322349","https://openalex.org/W2348545647","https://openalex.org/W2052976169","https://openalex.org/W2369600518","https://openalex.org/W2029144766","https://openalex.org/W2089275957","https://openalex.org/W2983681086","https://openalex.org/W2964031101"],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"have":[4],"achieved":[5],"significant":[6],"advances,":[7],"but":[8],"can":[9,83],"potentially":[10],"generate":[11],"harmful":[12,93],"content":[13],"such":[14],"as":[15],"social":[16],"biases,":[17],"extremism,":[18],"and":[19,36,106,119],"misinformation.Red":[20],"teaming":[21],"is":[22],"a":[23,99],"promising":[24],"approach":[25],"to":[26,34,90,92,132,153],"enhance":[27],"model":[28,38,72,101,130],"safety":[29],"by":[30],"generating":[31],"adversarial":[32,104,156,173],"prompts":[33],"test":[35],"improve":[37],"robustness.However,":[39],"existing":[40],"red-teaming":[41,61],"methods":[42],"often":[43],"require":[44],"expensive":[45],"fine-tuning,":[46],"especially":[47],"for":[48],"large":[49,128],"LLMs.We":[50],"propose":[51],"the":[52,70,76,86,108,113,124,127,133,138,150,164],"Dynamic":[53],"Evil":[54],"Score-Guided":[55],"Decoding":[56],"framework":[57,74],"(DESGD),":[58],"an":[59,79,103,145,168],"efficient":[60],"method":[62,143],"that":[63,141],"does":[64],"not":[65],"increase":[66],"computational":[67,161],"cost":[68],"with":[69,155],"target":[71,129],"size.DESGD":[73],"introduces":[75],"concept":[77],"of":[78,88,126,137,147,170],"'evil":[80],"score,'":[81],"which":[82],"dynamically":[84],"evaluate":[85],"potential":[87],"tokens":[89],"contribute":[91],"outputs":[94],"in":[95,115],"decoding":[96],"phrases.We":[97],"fine-tune":[98],"small":[100],"using":[102,159],"dataset":[105],"calculate":[107],"evil":[109,134],"score":[110],"based":[111],"on":[112,149,163],"difference":[114],"logits":[116,125],"vector":[117],"before":[118],"after":[120],"fine-tuning.Then,":[121],"we":[122],"adjust":[123],"according":[131],"score.The":[135],"results":[136],"experiment":[139],"show":[140],"our":[142],"achieves":[144],"ASR":[146,169],"92.83%":[148],"Llama-3.2-3B-Instructmodel,":[151],"compared":[152],"83.48%":[154],"fine-tuning":[157,174],"while":[158],"lower":[160],"resources.Similarly,":[162],"Qwen2.5-3B-Instructmodel,":[165],"DESGD":[166],"reaches":[167],"88.62%,":[171],"outperforming":[172],"(77.56%).":[175]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}