{"id":"https://openalex.org/W7201839693","doi":"https://doi.org/10.48550/arxiv.2608.05466","title":"Recursive Synthesis for Long-Horizon Terminal Tasks","display_name":"Recursive Synthesis for Long-Horizon Terminal Tasks","publication_year":2026,"publication_date":"2026-08-05","ids":{"openalex":"https://openalex.org/W7201839693","doi":"https://doi.org/10.48550/arxiv.2608.05466"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.05466","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.05466","pdf_url":"https://arxiv.org/pdf/2608.05466","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.05466","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5146047068","display_name":"Zhongzhi Li","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I165733156","display_name":"University of Georgia","ror":"https://ror.org/00te3t702","country_code":"US","type":"education","lineage":["https://openalex.org/I165733156"]},{"id":"https://openalex.org/I66946132","display_name":"University of Maryland, College Park","ror":"https://ror.org/047s2c258","country_code":"US","type":"education","lineage":["https://openalex.org/I66946132"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Li, Zhongzhi","raw_affiliation_strings":["University of Georgia","University of Maryland , College Park"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Georgia","institution_ids":["https://openalex.org/I165733156"]},{"raw_affiliation_string":"University of Maryland , College Park","institution_ids":["https://openalex.org/I66946132"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137809284","display_name":"Yucheng Shi","orcid":"https://orcid.org/0000-0002-8070-5363","observed_orcids":["https://orcid.org/0000-0002-8070-5363"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Yucheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140016173","display_name":"Zongxia Li","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I165733156","display_name":"University of Georgia","ror":"https://ror.org/00te3t702","country_code":"US","type":"education","lineage":["https://openalex.org/I165733156"]},{"id":"https://openalex.org/I66946132","display_name":"University of Maryland, College Park","ror":"https://ror.org/047s2c258","country_code":"US","type":"education","lineage":["https://openalex.org/I66946132"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Li, Zongxia","raw_affiliation_strings":["University of Georgia","University of Maryland , College Park"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Georgia","institution_ids":["https://openalex.org/I165733156"]},{"raw_affiliation_string":"University of Maryland , College Park","institution_ids":["https://openalex.org/I66946132"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075287872","display_name":"Ruhan Wang","orcid":"https://orcid.org/0000-0001-5874-4350","observed_orcids":["https://orcid.org/0000-0001-5874-4350"]},"institutions":[{"id":"https://openalex.org/I592451","display_name":"Indiana University","ror":"https://ror.org/01kg8sb98","country_code":"US","type":"education","lineage":["https://openalex.org/I592451"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wang, Ruhan","raw_affiliation_strings":["Indiana University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Indiana University","institution_ids":["https://openalex.org/I592451"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140946387","display_name":"Anhao Li","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I130238516","display_name":"University of Minnesota","ror":"https://ror.org/017zqws13","country_code":"US","type":"education","lineage":["https://openalex.org/I130238516"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Li, Anhao","raw_affiliation_strings":["University of Minnesota , Twin Cities"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Minnesota , Twin Cities","institution_ids":["https://openalex.org/I130238516"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5145958734","display_name":"Zixun Huang","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":"Huang, Zixun","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/A5146059838","display_name":"Junyao Yang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Yang, Junyao","raw_affiliation_strings":["National University of Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Singapore","institution_ids":["https://openalex.org/I165932596"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077165692","display_name":"Lei Ke","orcid":"https://orcid.org/0000-0002-0944-720X","observed_orcids":["https://orcid.org/0000-0002-0944-720X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ke, Lei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5145975243","display_name":"Ninghao Liu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I14243506","display_name":"Hong Kong Polytechnic University","ror":"https://ror.org/0030zas98","country_code":"HK","type":"education","lineage":["https://openalex.org/I14243506"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Liu, Ninghao","raw_affiliation_strings":["Hong Kong Polytechnic University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hong Kong Polytechnic University","institution_ids":["https://openalex.org/I14243506"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071533156","display_name":"Haitao Mi","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mi, Haitao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5140889669","display_name":"Leowei Liang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Leowei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":7,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.2078000009059906,"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.2078000009059906,"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":0.12300000339746475,"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/T13904","display_name":"Artificial Intelligence Applications","score":0.08579999953508377,"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/task-synthesis","display_name":"task synthesis","score":0.6899999976158142},{"id":"https://openalex.org/keywords/long-horizon-tasks","display_name":"long-horizon tasks","score":0.7110000252723694},{"id":"https://openalex.org/keywords/synthetic-training-data","display_name":"synthetic training data","score":0.5960000157356262},{"id":"https://openalex.org/keywords/agentic-reinforcement-learning","display_name":"agentic reinforcement learning","score":0.5389999747276306},{"id":"https://openalex.org/keywords/supervised-fine-tuning","display_name":"supervised fine-tuning","score":0.7070000171661377},{"id":"https://openalex.org/keywords/qwen3-5","display_name":"Qwen3.5","score":0.7509999871253967}],"concepts":[{"id":"https://openalex.org/C168773036","wikidata":"https://www.wikidata.org/wiki/Q264164","display_name":"Recursion (computer science)","level":2,"score":0.8046000003814697},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7465000152587891},{"id":"https://openalex.org/C2779664074","wikidata":"https://www.wikidata.org/wiki/Q3518405","display_name":"Terminal (telecommunication)","level":2,"score":0.7020999789237976},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.684499979019165},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.6019999980926514},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.4481000006198883},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.4020000100135803},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3873000144958496},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3560999929904938},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.3465000092983246},{"id":"https://openalex.org/C134121241","wikidata":"https://www.wikidata.org/wiki/Q899301","display_name":"Yield (engineering)","level":2,"score":0.3368000090122223},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.3203999996185303},{"id":"https://openalex.org/C2982939207","wikidata":"https://www.wikidata.org/wiki/Q1175101","display_name":"Recursive functions","level":2,"score":0.3100999891757965},{"id":"https://openalex.org/C164624739","wikidata":"https://www.wikidata.org/wiki/Q754331","display_name":"Activity-based costing","level":2,"score":0.3018999993801117},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.27810001373291016},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.27810001373291016}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.05466","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.05466","pdf_url":"https://arxiv.org/pdf/2608.05466","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.2608.05466","doi":"https://doi.org/10.48550/arxiv.2608.05466","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.05466","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:2608.05466","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.05466","pdf_url":"https://arxiv.org/pdf/2608.05466","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/W7201839693.pdf","grobid_xml":"https://content.openalex.org/works/W7201839693.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7171960007","https://openalex.org/W7221576374","https://openalex.org/W7165633594","https://openalex.org/W7203954397","https://openalex.org/W7128504532","https://openalex.org/W7161983523","https://openalex.org/W7158113780","https://openalex.org/W7168172066","https://openalex.org/W4415311645","https://openalex.org/W7162785433"],"abstract_inverted_index":{"High-quality":[0],"long-horizon":[1,65],"training":[2,161],"data":[3],"for":[4,63,102,175],"terminal":[5],"agents":[6],"is":[7],"expensive":[8],"to":[9,14,85,133,145,155,188,205,215],"produce,":[10],"often":[11,47],"costing":[12],"hundreds":[13],"thousands":[15],"of":[16,139,218],"dollars":[17],"per":[18,118],"task,":[19],"because":[20],"each":[21],"task":[22],"must":[23],"keep":[24],"the":[25,77,81,86,90,126,136,169,211,224,231,249,255],"instruction,":[26],"environment,":[27],"reference":[28,78,128],"solution,":[29,79],"and":[30,39,83,96,147,172,184,196,208,221,238],"verifier":[31,82],"mutually":[32],"consistent.":[33],"Human":[34],"authoring":[35],"does":[36],"not":[37],"scale,":[38],"direct":[40],"generation":[41],"with":[42],"large":[43],"language":[44],"models":[45],"(LLMs)":[46],"breaks":[48],"these":[49,180],"dependencies.":[50],"We":[51],"present":[52],"Recursive":[53],"Synthetic":[54],"Terminal":[55,198],"Tasks":[56],"(RST),":[57],"a":[58,93],"recursive":[59,107],"verified":[60,72],"synthesis":[61,236],"framework":[62],"constructing":[64],"terminal-agent":[66,113],"tasks":[67,99,114,171],"at":[68,115,153,157],"scale.":[69],"Starting":[70],"from":[71,131,143,151],"seed":[73],"tasks,":[74],"RST":[75,109],"extends":[76],"realigns":[80],"instruction":[84],"new":[87],"workflow,":[88],"validates":[89],"result":[91],"in":[92],"fresh":[94],"sandbox,":[95],"reuses":[97],"accepted":[98],"as":[100,243],"seeds":[101],"subsequent":[103],"rounds.":[104],"Across":[105],"fifteen":[106],"rounds,":[108,230],"produces":[110],"37,484":[111],"synthesized":[112,170],"roughly":[116],"$0.05":[117],"task.":[119],"Task":[120],"difficulty":[121,244],"increases":[122],"substantially":[123],"over":[124,223],"rounds:":[125],"median":[127,137],"solution":[129],"grows":[130,142],"67":[132],"374":[134],"lines,":[135],"number":[138],"executed":[140],"commands":[141],"40":[144],"244,":[146],"DeepSeek-V4-Pro":[148],"pass@4":[149],"drops":[150],"90%":[152],"R1":[154],"2.5%":[156],"R15.":[158],"To":[159],"demonstrate":[160],"utility,":[162],"we":[163],"collect":[164],"rejection-sampled":[165],"Qwen3.5":[166],"trajectories":[167,181],"on":[168,179,191,210],"use":[173],"them":[174],"supervised":[176],"fine-tuning.":[177],"Fine-tuning":[178],"improves":[182],"Qwen3.5-27B":[183,204],"Qwen3.5-122B-A10B":[185],"by":[186],"up":[187],"10":[189],"points":[190],"Terminal-Bench":[192,194],"2,":[193],"Hard,":[195],"Long-Horizon":[197],"Bench,":[199],"while":[200],"agentic":[201],"PPO":[202],"lifts":[203],"49.44%,":[206],"32.00%,":[207],"22.07%":[209],"three":[212],"benchmarks,":[213],"corresponding":[214],"relative":[216],"gains":[217],"20.0%,":[219],"41.2%,":[220],"21.9%":[222],"base":[225],"model.":[226],"Moreover,":[227],"after":[228],"15":[229],"recursion":[232],"shows":[233],"no":[234],"ceiling:":[235],"yield":[237],"validation":[239],"rates":[240],"remain":[241],"stable":[242],"keeps":[245],"climbing,":[246],"indicating":[247],"that":[248],"process":[250],"can":[251],"continue":[252],"well":[253],"beyond":[254],"scale":[256],"reported":[257],"here.":[258]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-08T00:00:00"}