{"id":"https://openalex.org/W4412889730","doi":"https://doi.org/10.18653/v1/2025.acl-long.1468","title":"PIGuard: Prompt Injection Guardrail via Mitigating Overdefense for Free","display_name":"PIGuard: Prompt Injection Guardrail via Mitigating Overdefense for Free","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412889730","doi":"https://doi.org/10.18653/v1/2025.acl-long.1468"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.1468","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1468","pdf_url":"https://aclanthology.org/2025.acl-long.1468.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 63rd Annual Meeting 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/2025.acl-long.1468.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100348666","display_name":"Hao Li","orcid":"https://orcid.org/0000-0002-9759-2537"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hao Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063002956","display_name":"Xiaogeng Liu","orcid":"https://orcid.org/0009-0008-7677-2310"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaogeng Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119181849","display_name":"Ning Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ning Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5005843046","display_name":"Chaowei Xiao","orcid":"https://orcid.org/0000-0002-7043-4926"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chaowei Xiao","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":3,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"30420","last_page":"30437"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T14445","display_name":"Transportation Safety and Impact Analysis","score":0.9950000047683716,"subfield":{"id":"https://openalex.org/subfields/2205","display_name":"Civil and Structural Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T14445","display_name":"Transportation Safety and Impact Analysis","score":0.9950000047683716,"subfield":{"id":"https://openalex.org/subfields/2205","display_name":"Civil and Structural Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.38100117444992065}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.38100117444992065}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.1468","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1468","pdf_url":"https://aclanthology.org/2025.acl-long.1468.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.1468","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1468","pdf_url":"https://aclanthology.org/2025.acl-long.1468.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412889730.pdf","grobid_xml":"https://content.openalex.org/works/W4412889730.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/W4396696052"],"abstract_inverted_index":{"Prompt":[0],"injection":[1,66,142],"attacks":[2],"pose":[3],"a":[4,94,101,134],"critical":[5],"threat":[6],"to":[7,33,84],"large":[8],"language":[9],"models":[10,75],"(LLMs),":[11],"enabling":[12,68],"goal":[13],"hijacking":[14],"and":[15,136],"data":[16],"leakage.Prompt":[17],"guard":[18,53,97],"models,":[19],"though":[20],"effective":[21],"in":[22,64],"defense,":[23],"suffer":[24,76],"from":[25,77],"over-defense-falsely":[26],"flagging":[27],"benign":[28,57],"inputs":[29],"as":[30],"malicious":[31],"due":[32],"trigger":[34,61,116],"word":[35],"bias.To":[36],"address":[37],"this":[38],"issue,":[39],"we":[40,91],"introduce":[41],"NotInject,":[42],"an":[43],"evaluation":[44],"dataset":[45],"that":[46,73,99],"systematically":[47],"measures":[48],"over-defense":[49,78],"across":[50],"various":[51],"prompt":[52,65,96,141],"models.NotInject":[54],"contains":[55],"339":[56],"samples":[58],"enriched":[59],"with":[60,80],"words":[62],"common":[63],"attacks,":[67],"fine-grained":[69],"evaluation.Our":[70],"results":[71],"show":[72],"state-of-the-art":[74,119],"issues,":[79],"accuracy":[81],"dropping":[82],"close":[83],"random":[85],"guessing":[86],"levels":[87],"(60%).To":[88],"mitigate":[89],"this,":[90],"propose":[92],"PIGuard,":[93],"novel":[95],"model":[98,130],"incorporates":[100],"new":[102],"training":[103],"strategy,":[104],"Mitigating":[105],"Over-defense":[106],"for":[107,139],"Free":[108],"(MOF),":[109],"which":[110],"significantly":[111],"reduces":[112],"the":[113,127],"bias":[114],"on":[115,121],"words.PIGuard":[117],"demonstrates":[118],"performance":[120],"diverse":[122],"benchmarks":[123],"including":[124],"Not-Inject,":[125],"surpassing":[126],"existing":[128],"best":[129],"by":[131],"30.4%,":[132],"offering":[133],"robust":[135],"open-source":[137],"solution":[138],"detecting":[140],"attacks.":[143]},"counts_by_year":[{"year":2026,"cited_by_count":3}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}