{"id":"https://openalex.org/W7137975510","doi":"https://doi.org/10.1609/aaai.v40i38.40543","title":"WALKSAFE: Risk-aware Graph Random Walk with Bi-GRPO for LLM Safety","display_name":"WALKSAFE: Risk-aware Graph Random Walk with Bi-GRPO for LLM Safety","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7137975510","doi":"https://doi.org/10.1609/aaai.v40i38.40543"},"language":"en","primary_location":{"id":"doi:10.1609/aaai.v40i38.40543","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i38.40543","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40543/44504","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40543/44504","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5129742523","display_name":"Shilong Pan","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shilong Pan","raw_affiliation_strings":["National University of Defense Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Defense Technology","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129657508","display_name":"Zhiliang Tian","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhiliang Tian","raw_affiliation_strings":["National University of Defense Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Defense Technology","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121242370","display_name":"Wanlong Yu","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wanlong Yu","raw_affiliation_strings":["National University of Defense Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Defense Technology","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129704117","display_name":"Zhen huang","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhen Huang","raw_affiliation_strings":["National University of Defense Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Defense Technology","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108303609","display_name":"Qingyu Qiu","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qingyu Qiu","raw_affiliation_strings":["National University of Defense Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Defense Technology","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129725414","display_name":"Zihan Chen","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zihan Chen","raw_affiliation_strings":["National University of Defense Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Defense Technology","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113327331","display_name":"Zhonghao Sun","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhonghao Sun","raw_affiliation_strings":["National University of Defense Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Defense Technology","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129733384","display_name":"Minlie Huang","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Minlie Huang","raw_affiliation_strings":["Tsinghua University, Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University, Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5129674356","display_name":"Dongsheng Li","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dongsheng Li","raw_affiliation_strings":["National University of Defense Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Defense Technology","institution_ids":["https://openalex.org/I170215575"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"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":0.16800378,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"38","first_page":"32655","last_page":"32663"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.2892000079154968,"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/T10028","display_name":"Topic Modeling","score":0.2892000079154968,"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/T11273","display_name":"Advanced Graph Neural Networks","score":0.26440000534057617,"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":0.12919999659061432,"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/limiting","display_name":"Limiting","score":0.5550000071525574},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5145999789237976},{"id":"https://openalex.org/keywords/generalizability-theory","display_name":"Generalizability theory","score":0.5133000016212463},{"id":"https://openalex.org/keywords/graph","display_name":"Graph","score":0.5066999793052673},{"id":"https://openalex.org/keywords/semantics","display_name":"Semantics (computer science)","score":0.4821999967098236},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.3100000023841858}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6741999983787537},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.5550000071525574},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5145999789237976},{"id":"https://openalex.org/C27158222","wikidata":"https://www.wikidata.org/wiki/Q5532422","display_name":"Generalizability theory","level":2,"score":0.5133000016212463},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.5066999793052673},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.4821999967098236},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.39419999718666077},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.3937999904155731},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.3100000023841858},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3095000088214874},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.30140000581741333},{"id":"https://openalex.org/C146380142","wikidata":"https://www.wikidata.org/wiki/Q1137726","display_name":"Directed graph","level":2,"score":0.2930999994277954},{"id":"https://openalex.org/C12174686","wikidata":"https://www.wikidata.org/wiki/Q1058438","display_name":"Risk assessment","level":2,"score":0.27570000290870667},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26570001244544983},{"id":"https://openalex.org/C32896092","wikidata":"https://www.wikidata.org/wiki/Q189447","display_name":"Risk management","level":2,"score":0.26499998569488525},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.2556000053882599}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1609/aaai.v40i38.40543","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i38.40543","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40543/44504","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},{"id":"pmh:oai:ojs.aaai.org:article/40543","is_oa":false,"landing_page_url":"https://ojs.aaai.org/index.php/AAAI/article/view/40543","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2159-5399","raw_type":"info:eu-repo/semantics/article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i38.40543","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i38.40543","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40543/44504","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","score":0.7639300227165222,"display_name":"Peace, Justice and strong institutions"}],"awards":[{"id":"https://openalex.org/G2277931485","display_name":null,"funder_award_id":"62421002","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G2367558620","display_name":null,"funder_award_id":"62125604","funder_id":"https://openalex.org/F4320336125","funder_display_name":"National Science Fund for Distinguished Young Scholars"},{"id":"https://openalex.org/G2511849395","display_name":null,"funder_award_id":"62376284","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3606055544","display_name":null,"funder_award_id":"62025208","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5729763073","display_name":null,"funder_award_id":"62325604","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320336125","display_name":"National Science Fund for Distinguished Young Scholars","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7137975510.pdf","grobid_xml":"https://content.openalex.org/works/W7137975510.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,299],"(LLMs)":[3],"may":[4],"generate":[5,183],"harmful":[6,33,70,100,115,172,230,281],"outputs":[7],"on":[8,21,52,211,293],"malicious":[9,54,95,124],"inputs.":[10,55,71],"Existing":[11],"safety":[12,40],"methods,":[13],"including":[14],"prompt":[15],"engineering":[16],"and":[17,87,101,106,108,112,138,162,181,187,195,221,239,265,282,285,289],"model":[18,89,192,278],"editing,":[19],"rely":[20],"hand-crafted":[22],"templates":[23],"or":[24,48],"target-driven":[25],"parameter":[26],"modifications,":[27],"limiting":[28],"their":[29,184],"generalizability":[30],"in":[31,41,164,197],"unseen":[32],"scenarios.":[34],"Post-training":[35],"aims":[36],"to":[37,64,76,132,154,177,214,232,236,250,279],"ensure":[38],"LLM":[39,235],"general":[42],"domains":[43],"via":[44,167],"supervised":[45],"fine-tuning":[46],"(SFT)":[47],"reinforcement":[49],"learning":[50],"(RL)":[51],"diverse":[53,69],"SFT":[56],"needs":[57],"annotated":[58],"refusal":[59],"samples":[60],"while":[61,117],"RL":[62],"learns":[63],"refuse":[65,78],"risk":[66,217],"by":[67,228],"exploring":[68],"However,":[72],"these":[73],"methods":[74],"tend":[75],"harshly":[77],"over":[79],"any":[80],"possible":[81],"risks,":[82],"sacrificing":[83],"potentially":[84],"useful":[85,119],"information":[86],"degrading":[88],"utility.":[90],"We":[91],"argue":[92],"that":[93,151,297],"realistic":[94],"inputs":[96,126,166,199],"often":[97],"mix":[98],"both":[99,263],"helpful":[102,188,290],"semantics":[103,160],"(i.e.,":[104],"entities":[105,194,220],"relations),":[107],"LLMs":[109,134,153,174,295],"should":[110],"identify":[111,155],"remove":[113],"only":[114,257],"relations":[116,196,231],"preserving":[118],"ones.":[120],"Thus,":[121],"the":[122,198,212,234,258,277],"original":[123],"user":[125,165],"can":[127,135,175],"shift":[128],"into":[129],"safe":[130,178,186,226,283,288],"queries,":[131],"which":[133],"respond":[136,176],"safely":[137,238],"helpfully.":[139],"In":[140],"this":[141],"paper,":[142],"we":[143,191,205,224,247],"propose":[144,248],"WALKSAFE,":[145],"a":[146,201,207],"graph-based":[147],"risk-aware":[148,208],"training":[149],"framework":[150],"enables":[152],"potential":[156,216],"risks":[157],"of":[158],"key":[159],"(entities":[161],"relations)":[163],"graph":[168,202,213],"structure.":[169,203],"By":[170],"filtering":[171,229],"relations,":[173],"input":[179],"queries":[180,227],"then":[182],"corresponding":[185],"responses.":[189,291],"First,":[190],"all":[193],"with":[200,243],"Second,":[204],"adopt":[206],"random":[209],"walk":[210],"quantify":[215],"under":[218],"multiple":[219],"relations.":[222],"Then,":[223],"reconstruct":[225],"promote":[233],"answer":[237],"helpfully":[240],"rather":[241],"than":[242],"direct":[244],"refusals.":[245],"Finally,":[246],"Bi-GRPO":[249,261],"post-train":[251],"LLMs.":[252],"As":[253],"vanilla":[254],"GRPO":[255],"conducts":[256],"intra-group":[259,264],"comparison,":[260],"performs":[262],"inter-group":[266,274],"comparisons":[267],"between":[268],"different":[269],"response":[270],"groups.":[271],"The":[272],"extra":[273],"rewards":[275],"encourage":[276],"distinguish":[280],"semantics,":[284],"thus":[286],"prefer":[287],"Experiments":[292],"three":[294],"show":[296],"our":[298],"obtain":[300],"SOTA":[301],"results.":[302]},"counts_by_year":[],"updated_date":"2026-07-02T09:51:11.867554","created_date":"2026-03-18T00:00:00"}