{"id":"https://openalex.org/W7147284941","doi":"https://doi.org/10.48550/arxiv.2603.28101","title":"Heddle: A Distributed Orchestration System for Agentic RL Rollout","display_name":"Heddle: A Distributed Orchestration System for Agentic RL Rollout","publication_year":2026,"publication_date":"2026-03-30","ids":{"openalex":"https://openalex.org/W7147284941","doi":"https://doi.org/10.48550/arxiv.2603.28101"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.28101","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.28101","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,"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.2603.28101","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5132580242","display_name":"Zili Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zili","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5000644890","display_name":"Yinmin Zhong","orcid":"https://orcid.org/0000-0002-2504-7652"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhong, Yinmin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132695431","display_name":"Chengxu Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Chengxu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132547113","display_name":"Chao Jin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jin, Chao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132572748","display_name":"Bingyang Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Bingyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Wei, Xinming","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Xinming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132684153","display_name":"Yuliang Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Yuliang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5132706325","display_name":"Xin Jin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jin, Xin","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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.474700003862381,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.474700003862381,"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/T10524","display_name":"Traffic control and management","score":0.04089999943971634,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10714","display_name":"Software-Defined Networks and 5G","score":0.03700000047683716,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/scheduling","display_name":"Scheduling (production processes)","score":0.6304000020027161},{"id":"https://openalex.org/keywords/trajectory","display_name":"Trajectory","score":0.5960999727249146},{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.550000011920929},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.534600019454956},{"id":"https://openalex.org/keywords/queueing-theory","display_name":"Queueing theory","score":0.5059000253677368},{"id":"https://openalex.org/keywords/orchestration","display_name":"Orchestration","score":0.5038999915122986},{"id":"https://openalex.org/keywords/idle","display_name":"Idle","score":0.39989998936653137}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7558000087738037},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.6304000020027161},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.5960999727249146},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.550000011920929},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.534600019454956},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.5175999999046326},{"id":"https://openalex.org/C22684755","wikidata":"https://www.wikidata.org/wiki/Q847526","display_name":"Queueing theory","level":2,"score":0.5059000253677368},{"id":"https://openalex.org/C199168358","wikidata":"https://www.wikidata.org/wiki/Q3367000","display_name":"Orchestration","level":3,"score":0.5038999915122986},{"id":"https://openalex.org/C16320812","wikidata":"https://www.wikidata.org/wiki/Q1812200","display_name":"Idle","level":2,"score":0.39989998936653137},{"id":"https://openalex.org/C160403385","wikidata":"https://www.wikidata.org/wiki/Q220543","display_name":"Queue","level":2,"score":0.3571999967098236},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.3481000065803528},{"id":"https://openalex.org/C2780609101","wikidata":"https://www.wikidata.org/wiki/Q17156588","display_name":"Resource management (computing)","level":2,"score":0.33379998803138733},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.31209999322891235},{"id":"https://openalex.org/C29202148","wikidata":"https://www.wikidata.org/wiki/Q287260","display_name":"Resource allocation","level":2,"score":0.29249998927116394},{"id":"https://openalex.org/C78766204","wikidata":"https://www.wikidata.org/wiki/Q555032","display_name":"Multi-core processor","level":2,"score":0.29019999504089355},{"id":"https://openalex.org/C3017813396","wikidata":"https://www.wikidata.org/wiki/Q17078173","display_name":"Resource constraints","level":2,"score":0.2745000123977661},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.26350000500679016},{"id":"https://openalex.org/C106189395","wikidata":"https://www.wikidata.org/wiki/Q176789","display_name":"Markov decision process","level":3,"score":0.2603999972343445},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.25360000133514404}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.28101","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.28101","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,"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.2603.28101","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.28101","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,"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":"Reduced inequalities","id":"https://metadata.un.org/sdg/10","score":0.5790534019470215}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Agentic":[0],"Reinforcement":[1],"Learning":[2],"(RL)":[3],"enables":[4],"LLMs":[5,32],"to":[6,78,102,120,132,163,170],"solve":[7],"complex":[8],"tasks":[9],"by":[10],"alternating":[11],"between":[12,31],"a":[13,18,75],"data-collection":[14],"rollout":[15,87,167],"phase":[16],"and":[17,33,68,83,99,112,123],"policy":[19],"training":[20],"phase.":[21],"During":[22],"rollout,":[23],"the":[24,80,134,158],"agent":[25],"generates":[26],"trajectories,":[27],"i.e.,":[28],"multi-step":[29],"interactions":[30],"external":[34],"tools.":[35],"Yet,":[36],"frequent":[37],"tool":[38,117],"calls":[39],"induce":[40],"long-tailed":[41],"trajectory":[42,54,62],"generation":[43],"that":[44,52,127,154],"bottlenecks":[45],"rollouts.":[46],"This":[47],"stems":[48],"from":[49],"step-centric":[50],"designs":[51],"ignore":[53],"context,":[55],"triggering":[56],"three":[57,91],"system":[58,77],"problems":[59],"for":[60,144],"long-tail":[61,138,159],"generation:":[63],"queueing":[64],"delays,":[65],"interference":[66],"overhead,":[67],"inflated":[69],"per-token":[70,135],"time.":[71],"We":[72],"propose":[73],"Heddle,":[74],"trajectory-centric":[76],"optimize":[79],"when,":[81],"where,":[82],"how":[84],"of":[85,137],"agentic":[86,150],"execution.":[88],"Heddle":[89,155],"integrates":[90],"core":[92],"mechanisms:":[93],"trajectory-level":[94],"scheduling":[95],"using":[96],"runtime":[97],"prediction":[98],"progressive":[100],"priority":[101],"minimize":[103,121],"cumulative":[104],"queueing;":[105],"trajectory-aware":[106],"placement":[107],"via":[108],"presorted":[109],"dynamic":[110],"programming":[111],"opportunistic":[113],"migration":[114],"during":[115],"idle":[116],"call":[118],"intervals":[119],"interference;":[122],"trajectory-adaptive":[124],"resource":[125],"manager":[126],"dynamically":[128],"tunes":[129],"model":[130],"parallelism":[131],"accelerate":[133],"time":[136],"trajectories":[139],"while":[140],"maintaining":[141],"high":[142],"throughput":[143,168],"short":[145],"trajectories.":[146],"Evaluations":[147],"across":[148],"diverse":[149],"RL":[151],"workloads":[152],"demonstrate":[153],"effectively":[156],"neutralizes":[157],"bottleneck,":[160],"achieving":[161],"up":[162],"2.5$\\times$":[164],"higher":[165],"end-to-end":[166],"compared":[169],"state-of-the-art":[171],"baselines.":[172]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-02T00:00:00"}