{"id":"https://openalex.org/W7161551695","doi":"https://doi.org/10.48550/arxiv.2605.15565","title":"AstraFlow: Dataflow-Oriented Reinforcement Learning for Agentic LLMs","display_name":"AstraFlow: Dataflow-Oriented Reinforcement Learning for Agentic LLMs","publication_year":2026,"publication_date":"2026-05-15","ids":{"openalex":"https://openalex.org/W7161551695","doi":"https://doi.org/10.48550/arxiv.2605.15565"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.15565","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15565","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.2605.15565","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136423400","display_name":"Haizhong Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Haizhong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136427188","display_name":"Yizhuo Di","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Di, Yizhuo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136437021","display_name":"Jiahui Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Jiahui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136368697","display_name":"Shuowei Jin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jin, Shuowei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5091840464","display_name":"Xiao Liu","orcid":"https://orcid.org/0000-0002-0405-5215"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Xueshen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136436738","display_name":"Yongji Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Yongji","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136388728","display_name":"Z. Morley Mao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mao, Z. Morley","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136448622","display_name":"Ion Stoica","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Stoica, Ion","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136450640","display_name":"Jiawei Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Jiawei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136397470","display_name":"Beidi Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Beidi","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/T10181","display_name":"Natural Language Processing Techniques","score":0.1899999976158142,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.1899999976158142,"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":0.11240000277757645,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.11169999837875366,"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/reinforcement-learning","display_name":"Reinforcement learning","score":0.7613999843597412},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.6341000199317932},{"id":"https://openalex.org/keywords/component","display_name":"Component (thermodynamics)","score":0.5504000186920166},{"id":"https://openalex.org/keywords/control","display_name":"Control (management)","score":0.5152000188827515},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.4609000086784363},{"id":"https://openalex.org/keywords/dataflow","display_name":"Dataflow","score":0.4043000042438507}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.7613999843597412},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7348999977111816},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.6341000199317932},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.5504000186920166},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.5152000188827515},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.4609000086784363},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.42649999260902405},{"id":"https://openalex.org/C96324660","wikidata":"https://www.wikidata.org/wiki/Q205446","display_name":"Dataflow","level":2,"score":0.4043000042438507},{"id":"https://openalex.org/C2778029271","wikidata":"https://www.wikidata.org/wiki/Q5421931","display_name":"Extension (predicate logic)","level":2,"score":0.4041000008583069},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.3418000042438507},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.32179999351501465},{"id":"https://openalex.org/C56739046","wikidata":"https://www.wikidata.org/wiki/Q192060","display_name":"Knowledge management","level":1,"score":0.28290000557899475},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.2630999982357025},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2549000084400177},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.2542000114917755}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.15565","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15565","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.2605.15565","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15565","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1],"(RL)":[2],"is":[3],"increasingly":[4],"used":[5],"to":[6,26,117],"improve":[7],"the":[8,72,115,143],"reasoning,":[9],"coding,":[10],"and":[11,42,71,107,125,138,154],"tool-use":[12],"capabilities":[13],"of":[14,52,74],"large":[15],"language":[16],"models,":[17],"but":[18,55],"agentic":[19,27,122],"RL":[20,25,48,78,90,123,174],"remains":[21],"prohibitively":[22],"expensive.":[23],"Scaling":[24],"LLMs":[28],"requires":[29,60],"supporting":[30],"complex":[31,120],"workloads,":[32,140],"including":[33],"multi-policy":[34,121,147,163],"collaborative":[35,164],"training,":[36,148,165],"while":[37,176],"efficiently":[38,126],"using":[39],"elastic,":[40],"heterogeneous,":[41],"cross-region":[43,152],"compute":[44,129],"resources.":[45,130],"Existing":[46],"LLM":[47],"systems":[49,175],"support":[50,119],"some":[51],"these":[53,83],"capabilities,":[54],"each":[56],"new":[57],"extension":[58],"often":[59],"dedicated":[61],"system":[62,79,91,116,145],"engineering.":[63],"This":[64],"burden":[65],"arises":[66],"from":[67],"trainer-centered":[68,95],"control":[69,96],"architectures":[70],"lack":[73],"principled":[75,98],"abstractions":[76],"for":[77],"components.":[80],"To":[81],"address":[82],"limitations,":[84],"we":[85],"propose":[86],"AstraFlow,":[87,102],"a":[88],"dataflow-oriented":[89],"that":[92,142],"replaces":[93],"conventional":[94],"with":[97],"component":[99],"abstractions.":[100],"In":[101,162],"rollout":[103],"services,":[104],"dataflow":[105],"management,":[106],"training":[108,179],"are":[109],"decoupled":[110],"into":[111],"autonomous":[112],"components,":[113],"enabling":[114],"natively":[118],"workloads":[124],"exploit":[127],"diverse":[128],"We":[131],"evaluate":[132],"AstraFlow":[133,166],"across":[134],"math,":[135],"code,":[136],"search,":[137],"AgentBench":[139],"showing":[141],"same":[144],"supports":[146],"elastic":[149],"scaling,":[150],"heterogeneous":[151],"execution,":[153],"composable":[155],"data":[156],"algorithms":[157],"without":[158],"system-level":[159],"code":[160],"changes.":[161],"achieves":[167],"comparable":[168],"or":[169],"better":[170],"accuracy":[171],"than":[172],"existing":[173],"speeding":[177],"up":[178],"time":[180],"by":[181],"2.7x.":[182]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-19T00:00:00"}