{"id":"https://openalex.org/W4416748856","doi":"https://doi.org/10.1109/iros60139.2025.11246761","title":"Refined Policy Distillation: From VLA Generalists to RL Experts","display_name":"Refined Policy Distillation: From VLA Generalists to RL Experts","publication_year":2025,"publication_date":"2025-10-19","ids":{"openalex":"https://openalex.org/W4416748856","doi":"https://doi.org/10.1109/iros60139.2025.11246761"},"language":null,"primary_location":{"id":"doi:10.1109/iros60139.2025.11246761","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iros60139.2025.11246761","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5094008630","display_name":"Tobias J\u00fclg","orcid":null},"institutions":[{"id":"https://openalex.org/I4210164862","display_name":"Artificial Intelligence in Medicine (Canada)","ror":"https://ror.org/05p590m36","country_code":"CA","type":"company","lineage":["https://openalex.org/I4210164862"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Tobias J\u00fclg","raw_affiliation_strings":["University of Technology,Department of Computer Science & Artificial Intelligence,Nuremberg,Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Technology,Department of Computer Science & Artificial Intelligence,Nuremberg,Germany","institution_ids":["https://openalex.org/I4210164862"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084499878","display_name":"Wolfram Burgard","orcid":"https://orcid.org/0000-0002-5680-6500"},"institutions":[{"id":"https://openalex.org/I4210164862","display_name":"Artificial Intelligence in Medicine (Canada)","ror":"https://ror.org/05p590m36","country_code":"CA","type":"company","lineage":["https://openalex.org/I4210164862"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Wolfram Burgard","raw_affiliation_strings":["University of Technology,Department of Computer Science & Artificial Intelligence,Nuremberg,Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Technology,Department of Computer Science & Artificial Intelligence,Nuremberg,Germany","institution_ids":["https://openalex.org/I4210164862"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5114402742","display_name":"Florian Walter","orcid":null},"institutions":[{"id":"https://openalex.org/I4210164862","display_name":"Artificial Intelligence in Medicine (Canada)","ror":"https://ror.org/05p590m36","country_code":"CA","type":"company","lineage":["https://openalex.org/I4210164862"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Florian Walter","raw_affiliation_strings":["University of Technology,Department of Computer Science & Artificial Intelligence,Nuremberg,Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Technology,Department of Computer Science & Artificial Intelligence,Nuremberg,Germany","institution_ids":["https://openalex.org/I4210164862"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I4210164862"],"apc_list":null,"apc_paid":null,"fwci":0.8823,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.80425847,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":null,"issue":null,"first_page":"11677","last_page":"11684"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.5023000240325928,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.5023000240325928,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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":0.2011999934911728,"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":0.1386999934911728,"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/reinforcement-learning","display_name":"Reinforcement learning","score":0.6909000277519226},{"id":"https://openalex.org/keywords/complement","display_name":"Complement (music)","score":0.6391000151634216},{"id":"https://openalex.org/keywords/policy-learning","display_name":"Policy learning","score":0.6092000007629395},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.6049000024795532},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.5809000134468079},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5594000220298767},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.5188000202178955}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6909000277519226},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6513000130653381},{"id":"https://openalex.org/C112313634","wikidata":"https://www.wikidata.org/wiki/Q7886648","display_name":"Complement (music)","level":5,"score":0.6391000151634216},{"id":"https://openalex.org/C2779436431","wikidata":"https://www.wikidata.org/wiki/Q30672407","display_name":"Policy learning","level":2,"score":0.6092000007629395},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.6049000024795532},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.5809000134468079},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5594000220298767},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5447999835014343},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5440999865531921},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5188000202178955},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.4323999881744385},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.39649999141693115},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.3553999960422516},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.30079999566078186},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2597000002861023},{"id":"https://openalex.org/C109986646","wikidata":"https://www.wikidata.org/wiki/Q546113","display_name":"Public policy","level":2,"score":0.25920000672340393},{"id":"https://openalex.org/C123587114","wikidata":"https://www.wikidata.org/wiki/Q2101508","display_name":"Policy analysis","level":2,"score":0.2574999928474426}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/iros60139.2025.11246761","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iros60139.2025.11246761","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321408","display_name":"Ministry of Education","ror":"https://ror.org/01p262204"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":9,"referenced_works":["https://openalex.org/W2158782408","https://openalex.org/W2963411833","https://openalex.org/W3035624836","https://openalex.org/W3088310808","https://openalex.org/W4385430679","https://openalex.org/W4401415287","https://openalex.org/W4401417065","https://openalex.org/W4402353985","https://openalex.org/W4414050916"],"related_works":[],"abstract_inverted_index":{"Vision-Language-Action":[0],"Models":[1],"(VLAs)":[2],"have":[3],"demonstrated":[4],"remarkable":[5],"generalization":[6],"capabilities":[7],"in":[8,94,117,139,165,188],"real-world":[9,140],"experiments.":[10],"However,":[11],"their":[12],"success":[13],"rates":[14],"are":[15,210],"often":[16],"not":[17],"on":[18,136],"par":[19],"with":[20,59],"expert":[21,76,158],"policies,":[22],"and":[23,70,98,110,168,191,208],"they":[24],"require":[25],"fine-tuning":[26],"when":[27],"the":[28,80,87,153,162,178,198],"setup":[29],"changes.":[30],"In":[31],"this":[32,50,120],"work,":[33],"we":[34],"introduce":[35],"Refined":[36],"Policy":[37],"Distillation":[38],"(RPD),":[39],"a":[40,54,90,122],"novel":[41],"Reinforcement":[42],"Learning":[43],"(RL)-based":[44],"policy":[45,82],"refinement":[46],"method":[47,104],"that":[48,150,160,197],"bridges":[49],"performance":[51,138],"gap":[52],"through":[53],"combination":[55],"of":[56,65,89,108],"on-policy":[57],"RL":[58,84,154,179],"behavioral":[60],"cloning.":[61],"The":[62],"core":[63],"idea":[64],"RPD":[66,116,151],"is":[67,121,183],"to":[68,114,156,186,194],"distill":[69],"refine":[71],"VLAs":[72],"into":[73],"compact,":[74],"high-performing":[75],"policies":[77,159],"by":[78,105],"guiding":[79],"student":[81,155],"during":[83],"exploration":[85],"using":[86],"actions":[88],"teacher":[91,164],"VLA,":[92],"resulting":[93],"increased":[95],"sample":[96],"efficiency":[97],"faster":[99,175],"convergence.":[100],"We":[101],"complement":[102],"our":[103],"fine-tuned":[106],"versions":[107],"Octo":[109],"OpenVLA":[111],"for":[112,125],"ManiSkill3":[113],"evaluate":[115],"simulation.":[118],"While":[119],"key":[123],"requirement":[124],"applying":[126],"RL,":[127],"it":[128],"also":[129,173],"yields":[130],"new":[131],"insights":[132],"beyond":[133],"existing":[134],"studies":[135],"VLA":[137,163,200,206],"settings.":[141],"Our":[142,181,203],"experimental":[143],"results":[144],"across":[145],"various":[146],"manipulation":[147],"tasks":[148],"show":[149],"enables":[152],"learn":[157],"outperform":[161],"both":[166],"dense":[167],"sparse":[169],"reward":[170],"settings,":[171],"while":[172],"achieving":[174],"convergence":[176],"than":[177],"baseline.":[180],"approach":[182],"even":[184],"robust":[185],"changes":[187],"camera":[189],"perspective":[190],"can":[192],"generalize":[193],"task":[195],"variations":[196],"underlying":[199],"cannot":[201],"solve.":[202],"code,":[204],"dataset,":[205],"checkpoints,":[207],"videos":[209],"available":[211],"at":[212],"https://refined-policy-distillation.github.io":[213]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-06-26T08:34:08.712188","created_date":"2025-11-28T00:00:00"}