{"id":"https://openalex.org/W4416034566","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.630","title":"Let Them Down Easy! Contextual Effects of LLM Guardrails on User Perceptions and Preferences","display_name":"Let Them Down Easy! Contextual Effects of LLM Guardrails on User Perceptions and Preferences","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416034566","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.630"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.630","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.630","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.630","pdf_url":"https://aclanthology.org/2025.findings-emnlp.630.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://aclanthology.org/2025.findings-emnlp.630.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5024099895","display_name":"Mingqian Zheng","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I74973139","display_name":"Carnegie Mellon University","ror":"https://ror.org/05x2bcf33","country_code":"US","type":"education","lineage":["https://openalex.org/I74973139"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mingqian Zheng","raw_affiliation_strings":["Carnegie Mellon University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Carnegie Mellon University","institution_ids":["https://openalex.org/I74973139"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104163221","display_name":"Wenjia Hu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210156221","display_name":"Allen Institute for Artificial Intelligence","ror":"https://ror.org/05w520734","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I4210156221"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wenjia Hu","raw_affiliation_strings":["University \u2663 Allen Institute for AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University \u2663 Allen Institute for AI","institution_ids":["https://openalex.org/I4210156221"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Patrick Zhao","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210156221","display_name":"Allen Institute for Artificial Intelligence","ror":"https://ror.org/05w520734","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I4210156221"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Patrick Zhao","raw_affiliation_strings":["University \u2663 Allen Institute for AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University \u2663 Allen Institute for AI","institution_ids":["https://openalex.org/I4210156221"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5054624773","display_name":"Motahhare Eslami","orcid":"https://orcid.org/0000-0002-1499-3045","observed_orcids":["https://orcid.org/0000-0002-1499-3045"]},"institutions":[{"id":"https://openalex.org/I74973139","display_name":"Carnegie Mellon University","ror":"https://ror.org/05x2bcf33","country_code":"US","type":"education","lineage":["https://openalex.org/I74973139"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Motahhare Eslami","raw_affiliation_strings":["Carnegie Mellon University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Carnegie Mellon University","institution_ids":["https://openalex.org/I74973139"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5080544237","display_name":"Jena D. Hwang","orcid":"https://orcid.org/0000-0003-3801-294X","observed_orcids":["https://orcid.org/0000-0003-3801-294X"]},"institutions":[{"id":"https://openalex.org/I4210156221","display_name":"Allen Institute for Artificial Intelligence","ror":"https://ror.org/05w520734","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I4210156221"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jena D. Hwang","raw_affiliation_strings":["University \u2663 Allen Institute for AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University \u2663 Allen Institute for AI","institution_ids":["https://openalex.org/I4210156221"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076297634","display_name":"Faeze Brahman","orcid":"https://orcid.org/0009-0008-4452-1488","observed_orcids":["https://orcid.org/0009-0008-4452-1488"]},"institutions":[{"id":"https://openalex.org/I74973139","display_name":"Carnegie Mellon University","ror":"https://ror.org/05x2bcf33","country_code":"US","type":"education","lineage":["https://openalex.org/I74973139"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Faeze Brahman","raw_affiliation_strings":["Carnegie Mellon University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Carnegie Mellon University","institution_ids":["https://openalex.org/I74973139"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089539629","display_name":"Carolyn Penstein Ros\u00e9","orcid":"https://orcid.org/0000-0003-1128-5155","observed_orcids":["https://orcid.org/0000-0003-1128-5155"]},"institutions":[{"id":"https://openalex.org/I4210156221","display_name":"Allen Institute for Artificial Intelligence","ror":"https://ror.org/05w520734","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I4210156221"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Carolyn Rose","raw_affiliation_strings":["University \u2663 Allen Institute for AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University \u2663 Allen Institute for AI","institution_ids":["https://openalex.org/I4210156221"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5015128745","display_name":"Maarten Sap","orcid":"https://orcid.org/0000-0002-0701-4654","observed_orcids":["https://orcid.org/0000-0002-0701-4654"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Maarten Sap","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":2.14,"has_fulltext":true,"cited_by_count":6,"citation_normalized_percentile":{"value":0.89445296,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":98,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"11742","last_page":"11772"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.6377000212669373,"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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.6377000212669373,"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":0.17739999294281006,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.06620000302791595,"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/refusal-strategies","display_name":"refusal strategies","score":0.8330000042915344},{"id":"https://openalex.org/keywords/partial-compliance","display_name":"partial compliance","score":0.800000011920929},{"id":"https://openalex.org/keywords/user-experience","display_name":"user experience","score":0.5059999823570251},{"id":"https://openalex.org/keywords/user-motivation","display_name":"user motivation","score":0.6190000176429749},{"id":"https://openalex.org/keywords/ai-safety","display_name":"AI safety","score":0.6679999828338623},{"id":"https://openalex.org/keywords/reward-models","display_name":"reward models","score":0.640999972820282},{"id":"https://openalex.org/keywords/user-perception","display_name":"user perception","score":0.3659999966621399},{"id":"https://openalex.org/keywords/user-engagement","display_name":"user engagement","score":0.3869999945163727}],"concepts":[{"id":"https://openalex.org/C26760741","wikidata":"https://www.wikidata.org/wiki/Q160402","display_name":"Perception","level":2,"score":0.5927000045776367},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.5702999830245972},{"id":"https://openalex.org/C77805123","wikidata":"https://www.wikidata.org/wiki/Q161272","display_name":"Social psychology","level":1,"score":0.5001999735832214},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.34209999442100525},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.32179999351501465},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.288100004196167},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.2655999958515167},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.2515000104904175},{"id":"https://openalex.org/C2776035688","wikidata":"https://www.wikidata.org/wiki/Q1606558","display_name":"Affect (linguistics)","level":2,"score":0.2515000104904175},{"id":"https://openalex.org/C75630572","wikidata":"https://www.wikidata.org/wiki/Q538904","display_name":"Applied psychology","level":1,"score":0.2305999994277954}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.630","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.630","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.630","pdf_url":"https://aclanthology.org/2025.findings-emnlp.630.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2506.00195","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2506.00195","pdf_url":"https://arxiv.org/pdf/2506.00195","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2506.00195","doi":"https://doi.org/10.48550/arxiv.2506.00195","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2506.00195","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.630","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.630","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.630","pdf_url":"https://aclanthology.org/2025.findings-emnlp.630.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G2258164063","display_name":"TITLE: CMU/NIST AI MEASUREMENT SCIENCE & ENGINEERING COOPERATIVE RESEARCH CENTERPURPOSE: ESTABLISH THE CMU/NIST AI MEASUREMENT SCIENCE & ENGINEERING COOPERATIVE RESEARCH CENTER (AIMSEC) TO FACILITATE COLLABORATIVE RESEARCH AND EXPERIMENTATION FOCUSED ON ADVANCING OUR NATIONAL CAPABILITY FOR TEST AND EVALUATION OF MODERN AI CAPABILITIES AND TOOLS. THERE ARE THREE PRINCIPAL GOALS: 1. ADVANCE MODERN AI RISK MANAGEMENT 2. VALIDATE EVALUATION APPROACHES THROUGH STAKEHOLDER PARTNERSHIPS. 3. TRANSLATE ASSESSMENT CAPABILITIES AND METHODOLOGIES TO PRACTICE. RESEARCH WILL FOCUS ON ADVANCING MEASUREMENT SCIENCE FOR MODERN AI SYSTEMS, INCLUDING MACHINE LEARNING (ML) AND GENERATIVE AI (GENAI) SYSTEMS, SUCH AS LARGE LANGUAGE MODELS (LLMS). ACTIVITIES TO BE PERFORMED: CREATE A PRACTICAL TAXONOMY OF THREATS AGAINST LARGE LANGUAGE MODELS (LLMS). THE PROPOSED WORK INCLUDES DEVELOPING TECHNIQUES TO REDUCE MEMORIZATION AND PROTECT CONFIDENTIAL INFORMATION PROVABLY, EVEN IF IT RESULTS IN LESS EXACT OUTPUTS. DEVELOP NEW MEASURES OF ROBUSTNESS FOR ML MODELS THAT BETTER REFLECT A MODEL?S ABILITY TO WITHSTAND ATTACKS IN PRACTICAL SCENARIOS. DEVELOP RICHER DEFINITIONS OF ALIGNMENT, CORRECTNESS, AND INTENT TO DETERMINE THE SUITABILITY OF GENERATIVE AI SYSTEMS FOR THEIR INTENDED USES. CREATE A STATE-OF-THE-ART SUITE OF ATTACKS AGAINST LLMS FOR TESTING NEW MODELS. THESE ACTIVITIES WILL BE ORGANIZED IN THREE STREAMS OF EFFORT ALIGNED WITH NIST AI PRIORITIES.  -\tVALIDITY, RELIABILITY, SAFETY, PRIVACY, AND SECURITY-\tACCOUNTABILITY, TRANSPARENCY, FAIRNESS, AND EXPLAINABILITY-\tGENERATIVE AI EVALUATIONEXPECTED OUTCOMES: DEVELOPMENT OF AI SYSTEM-LEVEL TOOLING, METRICS, EVALUATION PROCEDURES, DEVELOPMENT PROCESSES, AND BEST PRACTICES TO ENSURE AI SYSTEM BUILDERS CONSISTENTLY ENGINEER SAFE AI SYSTEMS. GOALS ARE TO ADVANCE MODERN AI RISK MANAGEMENT, VALIDATE EVALUATION APPROACHES THROUGH STAKEHOLDER PARTNERSHIPS, AND TRANSLATE ASSESSMENT CAPABILITIES AND METHODOLOGIES TO PRACTICE. RESEARCH WILL FOCUS ON ADVANCING MEASUREMENT SCIENCE FOR MODERN AI SYSTEMS, INCLUDING MACHINE LEARNING (ML) AND GENERATIVE AI (GENAI) SYSTEMS, SUCH AS LARGE LANGUAGE MODELS (LLMS).  THESE EFFORTS ARE EXPECTED TO YIELD ADDITIONAL APPROACHES TO MITIGATION OF WEAKNESSES AND VULNERABILITIES, PRIVACY PROTECTION, A PRACTICAL TAXONOMY OF THREATS AGAINST LLMS, APPROACHES TO MEET AI GOVERNANCE CHALLENGES, AND  NEW SOCIO-TECHNICAL DEFINITIONS AND METRICS.INTENDED BENEFICIARIES: AI STAKEHOLDERS, INCLUDING GOVERNMENT, INDUSTRY, ACADEMIA, AND THE GENERAL PUBLIC, WILL BENEFIT FROM THIS WORK OF ADVANCING CAPABILITY FOR TESTING AND EVALUATING MODERN AI CAPABILITIES AND TOOLS.SUBRECIPIENT ACTIVITIES: THERE ARE NO PLANNED SUBAWARDS.","funder_award_id":"60NANB24D231","funder_id":"https://openalex.org/F4320332178","funder_display_name":"National Institute of Standards and Technology"},{"id":"https://openalex.org/G6223126226","display_name":"Recrafting Computer Science: Fiber Crafting as Computational Thinking","funder_award_id":"2100401","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320310207","display_name":"Carnegie Mellon University","ror":"https://ror.org/05x2bcf33"},{"id":"https://openalex.org/F4320332178","display_name":"National Institute of Standards and Technology","ror":"https://ror.org/05xpvk416"},{"id":"https://openalex.org/F4320332180","display_name":"Defense Advanced Research Projects Agency","ror":"https://ror.org/02caytj08"},{"id":"https://openalex.org/F4320332815","display_name":"Advanced Research Projects Agency","ror":"https://ror.org/02caytj08"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416034566.pdf","grobid_xml":"https://content.openalex.org/works/W4416034566.grobid-xml"},"referenced_works_count":31,"referenced_works":["https://openalex.org/W3093233911","https://openalex.org/W4296413526","https://openalex.org/W4311991106","https://openalex.org/W4367623634","https://openalex.org/W4385571748","https://openalex.org/W4386794464","https://openalex.org/W4387561528","https://openalex.org/W4387725416","https://openalex.org/W4387838992","https://openalex.org/W4390833061","https://openalex.org/W4391591738","https://openalex.org/W4392297945","https://openalex.org/W4393063868","https://openalex.org/W4393905739","https://openalex.org/W4399317891","https://openalex.org/W4399991953","https://openalex.org/W4401042900","https://openalex.org/W4401306886","https://openalex.org/W4402671779","https://openalex.org/W4402672009","https://openalex.org/W4403160901","https://openalex.org/W4403702639","https://openalex.org/W4404306507","https://openalex.org/W4404330874","https://openalex.org/W4404782735","https://openalex.org/W4406032137","https://openalex.org/W4406879742","https://openalex.org/W4411119305","https://openalex.org/W4414942254","https://openalex.org/W4415230779","https://openalex.org/W6947695150"],"related_works":["https://openalex.org/W4414969592","https://openalex.org/W4396827165","https://openalex.org/W7160597081","https://openalex.org/W7213417091","https://openalex.org/W7204963031","https://openalex.org/W4412886755","https://openalex.org/W6910471513","https://openalex.org/W4416958115","https://openalex.org/W4416383393","https://openalex.org/W7213421399"],"abstract_inverted_index":{"Current":[0],"LLMs":[1,91],"are":[2],"trained":[3],"to":[4,80],"refuse":[5],"potentially":[6],"harmful":[7,16],"input":[8],"queries":[9],"regardless":[10],"of":[11,28,88],"whether":[12],"users":[13],"actually":[14],"had":[15],"intents,":[17],"causing":[18],"a":[19,26,132],"tradeoff":[20],"between":[21],"safety":[22,136,141],"and":[23,92,110,142],"user":[24,42,54,58,75,144],"experience.Through":[25],"study":[27],"480":[29],"participants":[30],"evaluating":[31],"3,840":[32],"query-response":[33],"pairs,":[34],"we":[35,84],"examine":[36],"how":[37,94],"different":[38,99],"refusal":[39,100],"strategies":[40],"affect":[41],"perceptions":[43,76],"across":[44],"varying":[45],"motivations.Our":[46],"findings":[47],"reveal":[48],"that":[49,103,118,138],"response":[50,86],"strategy":[51],"largely":[52],"shapes":[53],"experience,":[55],"while":[56],"actual":[57],"motivation":[59],"has":[60],"negligible":[61],"impact.Partial":[62],"complianceproviding":[63],"general":[64],"information":[65],"without":[66],"actionable":[67],"details-emerges":[68],"as":[69],"the":[70],"optimal":[71],"strategy,":[72],"reducing":[73],"negative":[74],"by":[77],"over":[78],"50%":[79],"flat-out":[81],"refusals.Complementing":[82],"this,":[83],"analyze":[85],"patterns":[87],"9":[89],"state-of-theart":[90],"evaluate":[93],"6":[95],"reward":[96,111],"models":[97,104,112],"score":[98],"strategies,":[101],"demonstrating":[102],"rarely":[105],"deploy":[106],"partial":[107],"compliance":[108],"naturally":[109],"currently":[113],"undervalue":[114],"it.This":[115],"work":[116],"demonstrates":[117],"effective":[119],"guardrails":[120],"require":[121],"focusing":[122],"on":[123],"crafting":[124],"thoughtful":[125],"refusals":[126],"rather":[127],"than":[128],"detecting":[129],"intent,":[130],"offering":[131],"path":[133],"toward":[134],"AI":[135],"mechanisms":[137],"ensure":[139],"both":[140],"sustained":[143],"engagement.":[145],"1":[146]},"counts_by_year":[{"year":2026,"cited_by_count":6}],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2025-11-08T00:00:00"}
