{"id":"https://openalex.org/W7224749986","doi":"https://doi.org/10.48550/arxiv.2610.39524","title":"Last-Iterate Convergence for Constrained MDPs: From Dual Descent to Primal-Dual","display_name":"Last-Iterate Convergence for Constrained MDPs: From Dual Descent to Primal-Dual","publication_year":2026,"publication_date":"2026-10-29","ids":{"openalex":"https://openalex.org/W7224749986","doi":"https://doi.org/10.48550/arxiv.2610.39524"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.39524","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39524","pdf_url":"https://arxiv.org/pdf/2610.39524","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/2610.39524","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5031262707","display_name":"Michael Lu","orcid":"https://orcid.org/0000-0003-4353-462X","observed_orcids":["https://orcid.org/0000-0003-4353-462X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu, Michael","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132678048","display_name":"Max Qiushi Lin","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Max Qiushi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160192640","display_name":"Mo Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I18014758","display_name":"Simon Fraser University","ror":"https://ror.org/0213rcc28","country_code":"CA","type":"education","lineage":["https://openalex.org/I18014758"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Chen, Mo","raw_affiliation_strings":["University Simon Fraser University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University Simon Fraser University","institution_ids":["https://openalex.org/I18014758"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5029028027","display_name":"Sharan Vaswani","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I18014758","display_name":"Simon Fraser University","ror":"https://ror.org/0213rcc28","country_code":"CA","type":"education","lineage":["https://openalex.org/I18014758"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Vaswani, Sharan","raw_affiliation_strings":["University Simon Fraser University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University Simon Fraser University","institution_ids":["https://openalex.org/I18014758"]}]}],"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":0.9976000189781189,"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":0.9976000189781189,"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/T12794","display_name":"Adaptive Dynamic Programming Control","score":0.0010000000474974513,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T12101","display_name":"Advanced Bandit Algorithms Research","score":0.0003000000142492354,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/constrained-markov-decision-process","display_name":"constrained Markov decision process","score":0.968999981880188},{"id":"https://openalex.org/keywords/last-iterate-convergence","display_name":"last-iterate convergence","score":0.9120000004768372},{"id":"https://openalex.org/keywords/primal-dual-algorithm","display_name":"primal-dual algorithm","score":0.8640000224113464},{"id":"https://openalex.org/keywords/policy-optimization","display_name":"policy optimization","score":0.7639999985694885},{"id":"https://openalex.org/keywords/dual-regularization","display_name":"dual regularization","score":0.9710000157356262},{"id":"https://openalex.org/keywords/tracking-analysis","display_name":"tracking analysis","score":0.7609999775886536},{"id":"https://openalex.org/keywords/bi-level-optimization","display_name":"bi-level optimization","score":0.8740000128746033},{"id":"https://openalex.org/keywords/linear-function-approximation","display_name":"linear function approximation","score":0.9129999876022339}],"concepts":[{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.6837999820709229},{"id":"https://openalex.org/C2780980858","wikidata":"https://www.wikidata.org/wiki/Q110022","display_name":"Dual (grammatical number)","level":2,"score":0.6133999824523926},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.6051999926567078},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.5023000240325928},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.47909998893737793},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.4481000006198883},{"id":"https://openalex.org/C115680565","wikidata":"https://www.wikidata.org/wiki/Q5977448","display_name":"Gradient method","level":2,"score":0.4410000145435333},{"id":"https://openalex.org/C106189395","wikidata":"https://www.wikidata.org/wiki/Q176789","display_name":"Markov decision process","level":3,"score":0.4316999912261963},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.42719998955726624},{"id":"https://openalex.org/C98763669","wikidata":"https://www.wikidata.org/wiki/Q176645","display_name":"Markov chain","level":2,"score":0.40700000524520874},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.39730000495910645},{"id":"https://openalex.org/C159886148","wikidata":"https://www.wikidata.org/wiki/Q176645","display_name":"Markov process","level":2,"score":0.3840999901294708},{"id":"https://openalex.org/C55660270","wikidata":"https://www.wikidata.org/wiki/Q5164377","display_name":"Constrained optimization","level":2,"score":0.3564000129699707},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.34630000591278076},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3395000100135803},{"id":"https://openalex.org/C2776637919","wikidata":"https://www.wikidata.org/wiki/Q624380","display_name":"Descent (aeronautics)","level":2,"score":0.32829999923706055},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.32820001244544983},{"id":"https://openalex.org/C3309286","wikidata":"https://www.wikidata.org/wiki/Q4907693","display_name":"Bilevel optimization","level":3,"score":0.31850001215934753},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.30649998784065247},{"id":"https://openalex.org/C41045048","wikidata":"https://www.wikidata.org/wiki/Q202843","display_name":"Linear programming","level":2,"score":0.30390000343322754},{"id":"https://openalex.org/C159694833","wikidata":"https://www.wikidata.org/wiki/Q2321565","display_name":"Iterative method","level":2,"score":0.2987000048160553},{"id":"https://openalex.org/C57869625","wikidata":"https://www.wikidata.org/wiki/Q1783502","display_name":"Rate of convergence","level":3,"score":0.29760000109672546},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.2822999954223633},{"id":"https://openalex.org/C116149140","wikidata":"https://www.wikidata.org/wiki/Q2070951","display_name":"Descent direction","level":4,"score":0.2709999978542328},{"id":"https://openalex.org/C2775936607","wikidata":"https://www.wikidata.org/wiki/Q466845","display_name":"Tracking (education)","level":2,"score":0.2615000009536743}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.39524","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39524","pdf_url":"https://arxiv.org/pdf/2610.39524","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.2610.39524","doi":"https://doi.org/10.48550/arxiv.2610.39524","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.39524","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:2610.39524","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39524","pdf_url":"https://arxiv.org/pdf/2610.39524","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/W7224749986.pdf","grobid_xml":"https://content.openalex.org/works/W7224749986.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7133188436","https://openalex.org/W7161019486","https://openalex.org/W4294691315","https://openalex.org/W4288874163","https://openalex.org/W4281783469","https://openalex.org/W4405426152","https://openalex.org/W4391462862","https://openalex.org/W4320086332","https://openalex.org/W4382318857","https://openalex.org/W6941246861"],"abstract_inverted_index":{"We":[0,93,114],"study":[1],"policy":[2,39,98],"optimization":[3],"for":[4,25],"infinite-horizon,":[5],"discounted":[6],"constrained":[7],"Markov":[8],"decision":[9],"processes.":[10],"While":[11],"primal-dual":[12],"($\\texttt{PD}$)":[13],"methods":[14,35,49,86],"are":[15,71],"widely":[16],"used":[17],"in":[18,55,75,109],"practice,":[19],"their":[20,59],"theoretical":[21,137],"guarantees":[22,43,172],"typically":[23],"hold":[24],"an":[26,56,110],"impractical":[27],"mixture":[28],"policy.":[29],"In":[30],"contrast,":[31],"dual":[32,101],"descent":[33],"($\\texttt{DD}$)":[34],"output":[36],"a":[37,81,116,143],"single":[38],"(\"last":[40],"iterate\")":[41],"with":[42],"on":[44,66],"its":[45],"performance.":[46],"However,":[47],"such":[48],"require":[50,90],"iteratively":[51],"solving":[52],"entropy-regularized":[53,96],"MDPs":[54],"inner-loop,":[57],"limiting":[58],"practical":[60],"effectiveness.":[61],"Moreover,":[62],"existing":[63,125],"analyses":[64],"rely":[65],"strong":[67],"coverage":[68,91],"assumptions":[69],"that":[70,87,147],"difficult":[72],"to":[73,103,124,141,159,165],"satisfy":[74],"practice.":[76],"Consequently,":[77],"we":[78,155,169],"first":[79],"provide":[80],"new":[82],"analysis":[83,119,158],"of":[84],"$\\texttt{DD}$":[85,150],"does":[88],"not":[89],"assumptions.":[92,176],"then":[94],"employ":[95],"natural":[97],"gradient":[99],"and":[100,151],"regularization":[102],"eliminate":[104],"the":[105,127],"inner":[106],"loop,":[107],"resulting":[108],"efficient":[111],"$\\texttt{PD}$":[112,152],"algorithm.":[113],"develop":[115],"novel":[117],"tracking":[118],"which":[120],"shows":[121],"that,":[122],"compared":[123],"methods,":[126,168],"proposed":[128],"algorithm":[129],"achieves":[130],"improved":[131],"last-iterate":[132,167],"convergence":[133],"guarantees.":[134],"Furthermore,":[135],"our":[136,157],"framework":[138],"allows":[139],"us":[140],"analyze":[142],"truncated":[144],"bilevel":[145],"method":[146],"interpolates":[148],"between":[149],"methods.":[153],"Finally,":[154],"generalize":[156],"handle":[160],"linear":[161],"function":[162],"approximation.":[163],"Compared":[164],"prior":[166],"prove":[170],"stronger":[171],"under":[173],"less":[174],"restrictive":[175]},"counts_by_year":[],"updated_date":"2026-11-02T09:45:03.842371","created_date":"2026-10-31T00:00:00"}