{"id":"https://openalex.org/W7138218903","doi":"https://doi.org/10.1609/aaai.v40i26.39278","title":"Behaviour Policy Optimization: Provably Lower Variance Return Estimates for Off-Policy Reinforcement Learning","display_name":"Behaviour Policy Optimization: Provably Lower Variance Return Estimates for Off-Policy Reinforcement Learning","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138218903","doi":"https://doi.org/10.1609/aaai.v40i26.39278"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i26.39278","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i26.39278","pdf_url":null,"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://doi.org/10.1609/aaai.v40i26.39278","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5129661535","display_name":"Alex W. Goodall","orcid":null},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":true,"raw_author_name":"Alex W. Goodall","raw_affiliation_strings":["Imperial College London"],"affiliations":[{"raw_affiliation_string":"Imperial College London","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129742763","display_name":"Edwin Hamel-De Le Court","orcid":null},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Edwin Hamel-De Le Court","raw_affiliation_strings":["Imperial College London"],"affiliations":[{"raw_affiliation_string":"Imperial College London","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5055883955","display_name":"Francesco Belardinelli","orcid":"https://orcid.org/0000-0002-7768-1794"},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Francesco Belardinelli","raw_affiliation_strings":["Imperial College London"],"affiliations":[{"raw_affiliation_string":"Imperial College London","institution_ids":["https://openalex.org/I47508984"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5129661535"],"corresponding_institution_ids":["https://openalex.org/I47508984"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.42987152,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"26","first_page":"21324","last_page":"21332"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.6474000215530396,"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":0.6474000215530396,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.03009999915957451,"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/T10845","display_name":"Advanced Causal Inference Techniques","score":0.029600000008940697,"subfield":{"id":"https://openalex.org/subfields/2613","display_name":"Statistics and Probability"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.7074000239372253},{"id":"https://openalex.org/keywords/variance","display_name":"Variance (accounting)","score":0.6733999848365784},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.6577000021934509},{"id":"https://openalex.org/keywords/sample","display_name":"Sample (material)","score":0.49059998989105225},{"id":"https://openalex.org/keywords/expected-return","display_name":"Expected return","score":0.4052000045776367},{"id":"https://openalex.org/keywords/reinforcement","display_name":"Reinforcement","score":0.4009000062942505}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.7074000239372253},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.6733999848365784},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.6577000021934509},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5723999738693237},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.49059998989105225},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.47200000286102295},{"id":"https://openalex.org/C154611145","wikidata":"https://www.wikidata.org/wiki/Q4331879","display_name":"Expected return","level":3,"score":0.4052000045776367},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.4009000062942505},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.3937000036239624},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.33410000801086426},{"id":"https://openalex.org/C58489278","wikidata":"https://www.wikidata.org/wiki/Q1172284","display_name":"Data set","level":2,"score":0.3208000063896179},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3192000091075897},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.2922999858856201},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.28949999809265137},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.26579999923706055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i26.39278","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i26.39278","pdf_url":null,"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"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i26.39278","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i26.39278","pdf_url":null,"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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Many":[0],"reinforcement":[1,82],"learning":[2,83],"algorithms,":[3],"particularly":[4],"those":[5],"that":[6,43],"rely":[7],"on":[8],"return":[9,26,58,174],"estimates":[10],"for":[11,54,112,167],"policy":[12,87,134,143,168],"improvement,":[13,169],"can":[14,47],"suffer":[15],"from":[16,35,127],"poor":[17],"sample":[18,189],"efficiency":[19,190],"and":[20,89,107,114,142,191],"training":[21],"instability":[22],"due":[23],"to":[24,50,79,93,164],"high-variance":[25],"estimates.":[27,59,175],"In":[28,176],"this":[29,76,185],"paper":[30],"we":[31,152,179],"leverage":[32],"new":[33],"results":[34],"off-policy":[36,52],"evaluation;":[37],"it":[38,65],"has":[39,99],"recently":[40],"been":[41,100],"shown":[42],"well-designed":[44],"behaviour":[45,159],"policies":[46],"be":[48],"used":[49,163],"collect":[51,165],"data":[53,68,125,166],"provably":[55,171],"lower":[56,172],"variance":[57,72,116,173],"This":[60],"result":[61],"is":[62,70,135,144,162],"surprising":[63],"as":[64],"means":[66],"collecting":[67],"on-policy":[69],"not":[71],"optimal.":[73],"We":[74],"extend":[75,180],"key":[77],"insight":[78],"the":[80,133,140,158],"online":[81],"setting,":[84],"where":[85],"both":[86],"evaluation":[88],"improvement":[90],"are":[91,121],"interleaved":[92],"learn":[94],"optimal":[95],"policies.":[96],"Off-policy":[97],"RL":[98],"well":[101],"studied":[102],"(e.g.,":[103],"IMPALA),":[104],"with":[105,123,170,184],"correct":[106],"truncated":[108],"importance":[109],"weighted":[110],"samples":[111],"de-biasing":[113],"managing":[115],"appropriately.":[117],"Generally":[118],"these":[119],"approaches":[120],"concerned":[122],"reconciling":[124],"collected":[126],"multiple":[128],"workers":[129,141],"in":[130,146],"parallel,":[131],"while":[132],"updated":[136],"asynchronously,":[137],"mismatch":[138],"between":[139],"corrected":[145],"a":[147,194],"mathematically":[148],"sound":[149],"way.":[150],"Here":[151],"consider":[153],"only":[154],"one":[155],"worker":[156],"-":[157],"policy,":[160],"which":[161],"our":[177],"experiments":[178],"two":[181],"policy-gradient":[182],"methods":[183],"regime,":[186],"demonstrating":[187],"better":[188],"performance":[192],"over":[193],"diverse":[195],"set":[196],"of":[197],"environments.":[198]},"counts_by_year":[],"updated_date":"2026-03-20T20:47:17.329874","created_date":"2026-03-18T00:00:00"}