{"id":"https://openalex.org/W7166843375","doi":"https://doi.org/10.18653/v1/2026.findings-acl.949","title":"Reverse Constitutional AI: A Framework for Controllable Toxic Data Generation via Probability-Clamped RLAIF","display_name":"Reverse Constitutional AI: A Framework for Controllable Toxic Data Generation via Probability-Clamped RLAIF","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166843375","doi":"https://doi.org/10.18653/v1/2026.findings-acl.949"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.949","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.949","pdf_url":"https://aclanthology.org/2026.findings-acl.949.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 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.949.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139803611","display_name":"Yuan Fang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuan Fang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139781526","display_name":"Yiming Luo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yiming Luo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139811795","display_name":"Aimin Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Aimin Zhou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139817255","display_name":"Fei Tan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fei Tan","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.83387865,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"19017","last_page":"19039"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.3368000090122223,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.3368000090122223,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.07930000126361847,"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/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.04540000110864639,"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/key","display_name":"Key (lock)","score":0.2667999863624573},{"id":"https://openalex.org/keywords/identification","display_name":"Identification (biology)","score":0.2529999911785126},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.24459999799728394},{"id":"https://openalex.org/keywords/data-collection","display_name":"Data collection","score":0.24289999902248383},{"id":"https://openalex.org/keywords/control","display_name":"Control (management)","score":0.2353000044822693}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5486999750137329},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2667999863624573},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2529999911785126},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.24459999799728394},{"id":"https://openalex.org/C133462117","wikidata":"https://www.wikidata.org/wiki/Q4929239","display_name":"Data collection","level":2,"score":0.24289999902248383},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.23899999260902405},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.2353000044822693},{"id":"https://openalex.org/C17500928","wikidata":"https://www.wikidata.org/wiki/Q959968","display_name":"Control system","level":2,"score":0.22669999301433563},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.22349999845027924},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.21870000660419464}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.949","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.949","pdf_url":"https://aclanthology.org/2026.findings-acl.949.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 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.949","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.949","pdf_url":"https://aclanthology.org/2026.findings-acl.949.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 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321881","display_name":"Shanghai Municipal Education Commission","ror":"https://ror.org/05tewj457"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166843375.pdf","grobid_xml":"https://content.openalex.org/works/W7166843375.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Ensuring":[0],"the":[1,13],"safety":[2,140],"of":[3,16,49,64,142],"large":[4],"language":[5,144],"models":[6],"(LLMs)":[7],"requires":[8],"robust":[9],"red":[10,134],"teaming,":[11],"yet":[12],"systematic":[14,139],"synthesis":[15,63],"high-quality":[17,111],"toxic":[18,112],"data":[19,34,67,113,136],"remains":[20],"under-explored.We":[21],"propose":[22],"Reverse":[23],"Constitutional":[24],"AI":[25,96],"(R-CAI),":[26],"a":[27,43,47,57,129],"framework":[28,132],"for":[29,72,133],"automated":[30,131],"and":[31,51,81,114,138],"controllable":[32],"adversarial":[33,66,100,104,125],"generation":[35,137],"that":[36,107,115],"moves":[37],"beyond":[38],"isolated":[39],"jailbreak":[40],"prompts.By":[41],"inverting":[42],"harmless":[44],"constitution":[45,48],"into":[46],"toxicity":[50],"iteratively":[52],"refining":[53],"model":[54],"outputs":[55],"through":[56],"critiquerevision":[58],"pipeline,":[59],"R-CAI":[60,108,127],"enables":[61],"scalable":[62],"multi-dimensional":[65],"without":[68,123],"human":[69],"annotation.Optimizing":[70],"solely":[71],"toxicity-related":[73],"rewards,":[74],"however,":[75],"can":[76],"lead":[77],"to":[78],"reward":[79],"hacking":[80],"degraded":[82],"semantic":[83,120],"coherence.To":[84],"address":[85],"this":[86],"challenge,":[87],"we":[88],"introduce":[89],"probability":[90,116],"clamping":[91,117],"within":[92],"reinforcement":[93],"learning":[94],"from":[95],"feedback,":[97],"which":[98],"stabilizes":[99],"optimization":[101],"while":[102],"preserving":[103],"intent.Experiments":[105],"demonstrate":[106],"generates":[109],"diverse,":[110],"substantially":[118],"improves":[119],"coherence":[121],"(15%)":[122],"sacrificing":[124],"strength.Overall,":[126],"provides":[128],"fully":[130],"teaming":[135],"evaluation":[141],"aligned":[143],"models.":[145]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}