{"id":"https://openalex.org/W4415328366","doi":"https://doi.org/10.48550/arxiv.2505.15062","title":"SAKE: Structured Agentic Knowledge Extrapolation for Complex LLM Reasoning via Reinforcement Learning","display_name":"SAKE: Structured Agentic Knowledge Extrapolation for Complex LLM Reasoning via Reinforcement Learning","publication_year":2025,"publication_date":"2025-05-21","ids":{"openalex":"https://openalex.org/W4415328366","doi":"https://doi.org/10.48550/arxiv.2505.15062"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2505.15062","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2505.15062","pdf_url":"https://arxiv.org/pdf/2505.15062","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2505.15062","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5024847539","display_name":"Jiashu He","orcid":"https://orcid.org/0009-0007-2616-5718","observed_orcids":["https://orcid.org/0009-0007-2616-5718"]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"He, Jiashu","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022812355","display_name":"Jin\u2010Xuan Fan","orcid":"https://orcid.org/0000-0002-8434-7591","observed_orcids":["https://orcid.org/0000-0002-8434-7591"]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Fan, Jinxuan","raw_affiliation_strings":["University of California , Berkeley"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Berkeley","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Jiang, Bowen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jiang, Bowen","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029632580","display_name":"Ignacio Hounie","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hounie, Ignacio","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023802054","display_name":"Dan Roth","orcid":"https://orcid.org/0009-0002-1447-5173","observed_orcids":["https://orcid.org/0009-0002-1447-5173"]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Roth, Dan","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5078862959","display_name":"Alejandro Ribeiro","orcid":"https://orcid.org/0000-0003-4230-9906","observed_orcids":["https://orcid.org/0000-0003-4230-9906"]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ribeiro, Alejandro","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":7,"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/T10028","display_name":"Topic Modeling","score":0.7949000000953674,"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":0.7949000000953674,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.06310000270605087,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T10286","display_name":"Information Retrieval and Search Behavior","score":0.04019999876618385,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/agentic-llms","display_name":"agentic LLMs","score":0.7110000252723694},{"id":"https://openalex.org/keywords/knowledge-graph-reasoning","display_name":"knowledge graph reasoning","score":0.8130000233650208},{"id":"https://openalex.org/keywords/associative-reasoning","display_name":"associative reasoning","score":0.8820000290870667},{"id":"https://openalex.org/keywords/analogical-reasoning","display_name":"analogical reasoning","score":0.6129999756813049},{"id":"https://openalex.org/keywords/group-relative-policy-optimization","display_name":"Group Relative Policy Optimization","score":0.6589999794960022},{"id":"https://openalex.org/keywords/biomedical-question-answering","display_name":"biomedical question answering","score":0.6159999966621399},{"id":"https://openalex.org/keywords/commonsense-question-answering","display_name":"commonsense question answering","score":0.921999990940094},{"id":"https://openalex.org/keywords/token-efficiency","display_name":"token efficiency","score":0.6740000247955322}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.777400016784668},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5838000178337097},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.48410001397132874},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4480000138282776},{"id":"https://openalex.org/C161301231","wikidata":"https://www.wikidata.org/wiki/Q3478658","display_name":"Knowledge representation and reasoning","level":2,"score":0.4250999987125397},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.42170000076293945},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.4205000102519989},{"id":"https://openalex.org/C159423971","wikidata":"https://www.wikidata.org/wiki/Q177251","display_name":"Associative property","level":2,"score":0.41659998893737793},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.4146000146865845},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.3686000108718872},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.3465999960899353},{"id":"https://openalex.org/C85407183","wikidata":"https://www.wikidata.org/wiki/Q1045785","display_name":"Semantic network","level":2,"score":0.34529998898506165},{"id":"https://openalex.org/C115925183","wikidata":"https://www.wikidata.org/wiki/Q1412694","display_name":"Knowledge-based systems","level":2,"score":0.3395000100135803},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.32690000534057617},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.325300008058548},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3206999897956848},{"id":"https://openalex.org/C37335422","wikidata":"https://www.wikidata.org/wiki/Q6888134","display_name":"Model-based reasoning","level":3,"score":0.31279999017715454},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.311599999666214},{"id":"https://openalex.org/C17305859","wikidata":"https://www.wikidata.org/wiki/Q382944","display_name":"Soar","level":2,"score":0.3009999990463257},{"id":"https://openalex.org/C118930307","wikidata":"https://www.wikidata.org/wiki/Q600590","display_name":"Tuple","level":2,"score":0.30090001225471497},{"id":"https://openalex.org/C4554734","wikidata":"https://www.wikidata.org/wiki/Q593744","display_name":"Knowledge base","level":2,"score":0.2976999878883362},{"id":"https://openalex.org/C195344581","wikidata":"https://www.wikidata.org/wiki/Q2555318","display_name":"Automated reasoning","level":2,"score":0.2793000042438507},{"id":"https://openalex.org/C30542707","wikidata":"https://www.wikidata.org/wiki/Q1603203","display_name":"Commonsense knowledge","level":3,"score":0.26969999074935913},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.2685000002384186},{"id":"https://openalex.org/C56289545","wikidata":"https://www.wikidata.org/wiki/Q6423376","display_name":"Knowledge integration","level":3,"score":0.26840001344680786},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.26019999384880066},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.25929999351501465}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2505.15062","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2505.15062","pdf_url":"https://arxiv.org/pdf/2505.15062","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2505.15062","doi":"https://doi.org/10.48550/arxiv.2505.15062","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2505.15062","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2505.15062","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2505.15062","pdf_url":"https://arxiv.org/pdf/2505.15062","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415328366.pdf","grobid_xml":"https://content.openalex.org/works/W4415328366.grobid-xml"},"referenced_works_count":14,"referenced_works":["https://openalex.org/W2113241217","https://openalex.org/W2252016937","https://openalex.org/W2898695519","https://openalex.org/W2964207259","https://openalex.org/W4361001766","https://openalex.org/W4387708435","https://openalex.org/W4395686609","https://openalex.org/W4403444190","https://openalex.org/W4403883066","https://openalex.org/W4415947132","https://openalex.org/W4416036988","https://openalex.org/W4417422257","https://openalex.org/W6891795438","https://openalex.org/W7091139669"],"related_works":["https://openalex.org/W4412945508","https://openalex.org/W7168399467","https://openalex.org/W7204940517","https://openalex.org/W4412944667","https://openalex.org/W4417517230","https://openalex.org/W4406611262","https://openalex.org/W4415251127","https://openalex.org/W4403443663","https://openalex.org/W6910555128","https://openalex.org/W4400341288"],"abstract_inverted_index":{"Knowledge":[0,41],"extrapolation":[1],"is":[2,16,20,34,107,208],"the":[3,123,200],"process":[4],"of":[5],"inferring":[6],"novel":[7],"information":[8],"by":[9,98,165,192],"combining":[10],"and":[11,54,71,95,119,128,155,203],"extending":[12],"existing":[13],"knowledge":[14,33,57,177],"that":[15,48,115,137,171],"explicitly":[17],"available.":[18],"It":[19],"essential":[21],"for":[22],"solving":[23],"complex":[24],"questions":[25],"in":[26],"specialized":[27],"domains":[28],"where":[29],"retrieving":[30],"comprehensive":[31],"external":[32,65],"impractical.":[35],"We":[36],"propose":[37],"SAKE":[38,62,138],"(Structured":[39],"Agentic":[40],"Extrapolation),":[42],"a":[43,85,113],"RL":[44],"powered":[45],"agentic":[46,146],"framework":[47],"trains":[49],"LLMs":[50],"to":[51,78,126,130],"autonomously":[52],"retrieve":[53,127],"extrapolate":[55],"structured":[56,176],"through":[58,102,196],"tool-augmented":[59],"reinforcement":[60,197],"learning.":[61],"defines":[63],"two":[64],"KG":[66,147],"tools:":[67],"entity":[68],"group":[69],"construction":[70],"cross-group":[72],"triplet":[73],"retrieval.":[74],"The":[75,104],"model":[76,124,141],"learns":[77],"interleave":[79],"these":[80],"2":[81],"retrieval":[82],"tools":[83,202],"during":[84],"three-turn":[86],"rollout:":[87],"extracting":[88],"key":[89],"entities,":[90],"filtering":[91],"relevant":[92],"concept":[93],"groups,":[94],"associative":[96,172],"reasoning":[97,148,173],"constructing":[99],"new":[100],"triplets":[101],"analogy.":[103],"entire":[105],"pipeline":[106],"optimized":[108],"end-to-end":[109,191],"with":[110,144,183,199],"GRPO":[111],"using":[112],"reward":[114],"combines":[116],"output":[117],"format":[118],"answer":[120],"correctness,":[121],"teaching":[122],"what":[125],"how":[129],"reason":[131],"over":[132,166,174],"it.":[133],"Our":[134,206],"experiments":[135],"proved":[136],"fine-tuned":[139],"Qwen2.5-7B":[140],"surpasses":[142],"GPT-3.5-Turbo":[143],"state-of-the-art":[145],"on":[149],"both":[150],"biomedical":[151],"(75.4%":[152],"vs.":[153,158],"70.1%)":[154],"commonsense":[156],"(81.3%":[157],"74.7%)":[159],"benchmarks,":[160],"while":[161],"reducing":[162],"token":[163],"usage":[164],"90%.":[167],"These":[168],"results":[169],"demonstrate":[170],"incomplete":[175],"does":[178],"not":[179],"require":[180],"large":[181],"models":[182,195],"complex,":[184],"multi-step":[185],"prompting,":[186],"thus":[187],"can":[188],"be":[189],"learned":[190],"small,":[193],"open-weight":[194],"learning":[198],"right":[201],"training":[204],"signal.":[205],"code":[207],"available":[209],"at":[210],"https://github.com/jxfan99/SAKE.":[211]},"counts_by_year":[{"year":2026,"cited_by_count":3},{"year":2025,"cited_by_count":4}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2025-10-19T00:00:00"}