{"id":"https://openalex.org/W7225180200","doi":"https://doi.org/10.48550/arxiv.2611.02609","title":"Whitened Advantage Summation Policy Optimization: Correlation-Aware Aggregation for Multi-Reward Reinforcement Learning","display_name":"Whitened Advantage Summation Policy Optimization: Correlation-Aware Aggregation for Multi-Reward Reinforcement Learning","publication_year":2026,"publication_date":"2026-11-02","ids":{"openalex":"https://openalex.org/W7225180200","doi":"https://doi.org/10.48550/arxiv.2611.02609"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2611.02609","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2611.02609","pdf_url":"https://arxiv.org/pdf/2611.02609","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":null,"license_id":null,"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/2611.02609","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5143720058","display_name":"Olaitan Olaleye","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I97018004","display_name":"Stanford University","ror":"https://ror.org/00f54p054","country_code":"US","type":"education","lineage":["https://openalex.org/I97018004"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Olaleye, Olaitan","raw_affiliation_strings":["ProofMed AI and Stanford University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"ProofMed AI and Stanford University","institution_ids":["https://openalex.org/I97018004"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5117691907","display_name":"Jayson Meribe","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I97018004","display_name":"Stanford University","ror":"https://ror.org/00f54p054","country_code":"US","type":"education","lineage":["https://openalex.org/I97018004"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Meribe, Jayson","raw_affiliation_strings":["ProofMed AI and Stanford University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"ProofMed AI and Stanford University","institution_ids":["https://openalex.org/I97018004"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5160560553","display_name":"Temidayo Olaleye","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Olaleye, Temidayo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.989300012588501,"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":0.989300012588501,"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":0.0015999999595806003,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.00139999995008111,"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/tool-calling","display_name":"tool calling","score":0.7699999809265137},{"id":"https://openalex.org/keywords/language-model-fine-tuning","display_name":"language model fine-tuning","score":0.6539999842643738}],"concepts":[{"id":"https://openalex.org/C136886441","wikidata":"https://www.wikidata.org/wiki/Q926129","display_name":"Normalization (sociology)","level":2,"score":0.8192999958992004},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.6420999765396118},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.5389999747276306},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.46549999713897705},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43540000915527344},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.43160000443458557},{"id":"https://openalex.org/C117220453","wikidata":"https://www.wikidata.org/wiki/Q5172842","display_name":"Correlation","level":2,"score":0.4203000068664551},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.3995000123977661},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.34880000352859497},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.31299999356269836},{"id":"https://openalex.org/C132964779","wikidata":"https://www.wikidata.org/wiki/Q2110223","display_name":"Raw data","level":2,"score":0.27649998664855957},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.26899999380111694},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26840001344680786},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.26829999685287476}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2611.02609","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2611.02609","pdf_url":"https://arxiv.org/pdf/2611.02609","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2611.02609","doi":"https://doi.org/10.48550/arxiv.2611.02609","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2611.02609","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:2611.02609","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2611.02609","pdf_url":"https://arxiv.org/pdf/2611.02609","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":null,"license_id":null,"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/W7225180200.pdf","grobid_xml":"https://content.openalex.org/works/W7225180200.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7214986218","https://openalex.org/W7216070695","https://openalex.org/W7120177137","https://openalex.org/W7161108613","https://openalex.org/W7203667350","https://openalex.org/W7164047897","https://openalex.org/W7222599721","https://openalex.org/W7127187585","https://openalex.org/W7162678769","https://openalex.org/W7166823600"],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1],"for":[2,31],"language":[3],"models":[4],"often":[5],"combines":[6],"several":[7],"rewards":[8,21,111,160,211,257],"by":[9,261],"normalizing":[10],"each":[11],"one":[12],"and":[13,127,163,180,214,234],"adding":[14,78],"them":[15],"with":[16,75,143,158,167,177,188,209,220],"fixed":[17],"weights.":[18],"When":[19],"the":[20,24,40,44,47,56,63,72,82,91,109,122,133,185,241,255,266,270],"are":[22,101],"correlated,":[23],"variation":[25],"they":[26],"share":[27],"is":[28,62,125],"counted":[29],"once":[30],"every":[32],"reward":[33,45,64,124,151],"that":[34,81,105,240],"carries":[35],"it.":[36],"In":[37,250],"terms":[38,219],"of":[39,43],"ZCA-decorrelated":[41],"components":[42],"signal,":[46],"weighted":[48],"sum":[49],"applies":[50],"weights":[51,58,84,107],"$\u03a3^{1/2}":[52],"w$":[53],"rather":[54,247],"than":[55,152,207,248],"chosen":[57,83],"$w$,":[59],"where":[60,254],"$\u03a3$":[61,137],"correlation":[65],"matrix.":[66],"We":[67,89],"study":[68],"WAS-PO,":[69],"which":[70,121,203],"whitens":[71],"normalized":[73],"advantages":[74],"$\u03a3^{-1/2}$":[76],"before":[77],"them,":[79],"so":[80],"apply":[85],"to":[86,200],"those":[87],"components.":[88],"derive":[90],"scalarized":[92],"objective":[93],"whose":[94],"gradient":[95],"it":[96],"follows":[97],"when":[98],"normalization":[99,176,246],"statistics":[100],"held":[102],"fixed,":[103],"show":[104],"its":[106],"on":[108,132],"raw":[110],"can":[112],"be":[113],"negative,":[114],"give":[115,128],"an":[116],"exact":[117],"two-reward":[118],"condition":[119,272],"under":[120],"lighter-weighted":[123,267],"reversed,":[126],"a":[129,147,235,251],"loose":[130],"bound":[131],"error":[134],"from":[135,244],"estimating":[136],"during":[138],"training.":[139],"On":[140],"tool":[141],"calling":[142],"Qwen2.5-1.5B-Instruct,":[144],"WAS-PO":[145,171,228,263],"reached":[146],"higher":[148],"mean":[149],"training":[150,232],"GDPO":[153,230],"in":[154,217,231],"all":[155,164],"seven":[156],"seeds":[157,166,223],"three":[159,165,189,210],"(3.8%":[161],"higher)":[162],"four":[168,191,221],"(3.9%":[169],"higher).":[170],"also":[172],"replaces":[173],"GDPO's":[174,208],"per-prompt":[175],"batch-level":[178,245],"normalization,":[179],"we":[181],"did":[182,196],"not":[183,197],"separate":[184],"two":[186,226,256],"changes":[187],"or":[190],"rewards.":[192],"The":[193],"training-reward":[194],"gains":[195],"carry":[198],"over":[199],"held-out":[201],"accuracy,":[202],"was":[204],"slightly":[205],"lower":[206,216],"(not":[212],"significant)":[213],"0.03":[215],"absolute":[218],"(two":[222],"evaluated).":[224],"With":[225],"rewards,":[227],"trailed":[229],"reward,":[233,268],"three-seed":[236],"control":[237],"weakly":[238],"suggests":[239],"gap":[242],"comes":[243],"whitening.":[249],"math":[252],"setting":[253],"were":[258],"nearly":[259],"identical":[260],"construction,":[262],"drove":[264],"down":[265],"as":[269],"sign":[271],"predicts.":[273]},"counts_by_year":[],"updated_date":"2026-11-06T12:05:46.655746","created_date":"2026-11-04T00:00:00"}