{"id":"https://openalex.org/W7163716327","doi":"https://doi.org/10.48550/arxiv.2606.05805","title":"From Risk Classification to Action Plan Remediation: A Guardrail Feedback Driven Framework for LLM Agents","display_name":"From Risk Classification to Action Plan Remediation: A Guardrail Feedback Driven Framework for LLM Agents","publication_year":2026,"publication_date":"2026-06-04","ids":{"openalex":"https://openalex.org/W7163716327","doi":"https://doi.org/10.48550/arxiv.2606.05805"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.05805","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05805","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.05805","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137989586","display_name":"Yuhao Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Yuhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138011718","display_name":"Jiacheng Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Jiacheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135224732","display_name":"Shaanan Cohney","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cohney, Shaanan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137935164","display_name":"Zhexin Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zhexin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137925704","display_name":"Feng Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Feng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137991142","display_name":"Xingliang Yuan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuan, Xingliang","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":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.7210000157356262,"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":0.7210000157356262,"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":0.08370000123977661,"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/T10028","display_name":"Topic Modeling","score":0.02850000001490116,"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/safer","display_name":"SAFER","score":0.7773000001907349},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.6486999988555908},{"id":"https://openalex.org/keywords/blocking","display_name":"Blocking (statistics)","score":0.4984999895095825},{"id":"https://openalex.org/keywords/plan","display_name":"Plan (archaeology)","score":0.4758000075817108},{"id":"https://openalex.org/keywords/action","display_name":"Action (physics)","score":0.42800000309944153},{"id":"https://openalex.org/keywords/safeguard","display_name":"Safeguard","score":0.40720000863075256},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.4009999930858612},{"id":"https://openalex.org/keywords/action-plan","display_name":"Action plan","score":0.3555000126361847}],"concepts":[{"id":"https://openalex.org/C2776654903","wikidata":"https://www.wikidata.org/wiki/Q2601463","display_name":"SAFER","level":2,"score":0.7773000001907349},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6486999988555908},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.5913000106811523},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5254999995231628},{"id":"https://openalex.org/C144745244","wikidata":"https://www.wikidata.org/wiki/Q4927286","display_name":"Blocking (statistics)","level":2,"score":0.4984999895095825},{"id":"https://openalex.org/C2776505523","wikidata":"https://www.wikidata.org/wiki/Q4785468","display_name":"Plan (archaeology)","level":2,"score":0.4758000075817108},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.42800000309944153},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.41850000619888306},{"id":"https://openalex.org/C2780771206","wikidata":"https://www.wikidata.org/wiki/Q3271761","display_name":"Safeguard","level":2,"score":0.40720000863075256},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.4009999930858612},{"id":"https://openalex.org/C2780210234","wikidata":"https://www.wikidata.org/wiki/Q422638","display_name":"Action plan","level":2,"score":0.3555000126361847},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.328000009059906},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.328000009059906},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.30390000343322754},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.28760001063346863},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.2720000147819519},{"id":"https://openalex.org/C186886427","wikidata":"https://www.wikidata.org/wiki/Q5441213","display_name":"Feedback loop","level":2,"score":0.2621999979019165},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.25850000977516174},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.25060001015663147},{"id":"https://openalex.org/C2778012447","wikidata":"https://www.wikidata.org/wiki/Q1034415","display_name":"Scope (computer science)","level":2,"score":0.25029999017715454}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.05805","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05805","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.05805","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05805","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16","score":0.5866062045097351}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"LLM-based":[0],"guardrails":[1,52,75],"typically":[2],"safeguard":[3],"agents":[4],"by":[5,41],"evaluating":[6],"proposed":[7],"actions":[8],"or":[9,47,144,155],"inputs":[10],"before":[11],"execution,":[12,157],"producing":[13],"safety":[14],"signals":[15],"such":[16],"as":[17,59,110],"binary":[18],"allow/deny":[19],"decisions,":[20],"risk":[21],"categories,":[22],"and/or":[23],"explanatory":[24],"rationales":[25],"about":[26],"potential":[27],"policy":[28],"violations.":[29],"However,":[30],"agent":[31,87,103,117,161,197],"risks":[32],"often":[33,53],"arise":[34],"when":[35],"otherwise":[36],"benign":[37,68,120,172],"tasks":[38],"are":[39],"contaminated":[40],"untrusted":[42],"external":[43],"content,":[44],"unsafe":[45],"instructions,":[46],"risky":[48],"tool":[49],"use.":[50],"Existing":[51],"flag":[54],"the":[55,63,67,116,160,171,181,209,218],"entire":[56],"task":[57,173],"uniformly":[58],"unsafe,":[60],"thereby":[61],"blocking":[62,156],"threat":[64],"but":[65],"sacrificing":[66],"part.":[69],"Moreover,":[70],"existing":[71],"work":[72],"largely":[73],"evaluates":[74],"in":[76],"isolation,":[77],"leaving":[78],"unclear":[79],"whether":[80],"their":[81],"interventions":[82],"lead":[83],"to":[84,114,136,162,214],"safer":[85],"downstream":[86],"behavior.":[88],"To":[89],"address":[90],"this,":[91],"we":[92],"introduce":[93],"TRIAD":[94,176,207],"(Tripartite":[95],"Response":[96],"for":[97],"Iterative":[98],"Agent":[99],"Guardrailing),":[100],"a":[101,111,128,132,190],"guardrail-integrated":[102,223],"framework":[104],"that":[105,206],"leverages":[106],"guardrail-generated":[107],"verbal":[108],"feedback":[109,179,195],"guiding":[112],"signal":[113],"keep":[115],"aligned":[118],"with":[119,147],"objectives":[121],"at":[122],"each":[123],"planning":[124],"step.":[125],"We":[126],"finetune":[127],"language":[129],"model":[130],"on":[131,201],"self-curated":[133],"training":[134],"dataset":[135],"output":[137],"one":[138],"of":[139],"three":[140],"decisions:":[141],"proceed,":[142],"refuse,":[143],"update,":[145],"together":[146],"structured":[148],"natural-language":[149],"feedback.":[150],"Rather":[151],"than":[152],"merely":[153],"allowing":[154],"update":[158],"guides":[159],"revise":[163],"its":[164],"plan,":[165],"avoid":[166],"harmful":[167],"components,":[168],"and":[169,188,196,203],"preserve":[170],"where":[174],"possible.":[175],"injects":[177],"this":[178],"into":[180],"agent's":[182],"context,":[183],"enabling":[184],"subsequent":[185],"plan":[186],"revision":[187],"forming":[189],"closed":[191],"loop":[192],"between":[193],"guardrail":[194],"planning.":[198],"Extensive":[199],"experiments":[200],"ASB":[202],"AgentHarm":[204],"show":[205],"reduces":[208],"average":[210],"attack":[211],"success":[212],"rate":[213],"10.42%,":[215],"while":[216],"achieving":[217],"best":[219],"safety-utility":[220],"trade-off":[221],"among":[222],"baselines.":[224],"Our":[225],"code":[226],"is":[227],"available":[228],"at:":[229],"https://github.com/YUHAOSUNABC/TRIAD.":[230]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-06T00:00:00"}