{"id":"https://openalex.org/W7166196117","doi":"https://doi.org/10.48550/arxiv.2606.26801","title":"Improving Vision-Language-Action Model Fine-Tuning with Structured Stage and Keyframe Supervision","display_name":"Improving Vision-Language-Action Model Fine-Tuning with Structured Stage and Keyframe Supervision","publication_year":2026,"publication_date":"2026-06-25","ids":{"openalex":"https://openalex.org/W7166196117","doi":"https://doi.org/10.48550/arxiv.2606.26801"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.26801","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.26801","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.26801","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5027459729","display_name":"Yan Xu","orcid":"https://orcid.org/0000-0003-0379-8866"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139458849","display_name":"Yixiang Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Yixiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139438530","display_name":"Kai Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Kai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139447483","display_name":"Jiabing Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Jiabing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139406769","display_name":"Peiyan Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Peiyan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109597339","display_name":"Q L","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Qisen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126420301","display_name":"Y Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139438318","display_name":"Liang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Liang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"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/T10653","display_name":"Robot Manipulation and Learning","score":0.8173999786376953,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10653","display_name":"Robot Manipulation and Learning","score":0.8173999786376953,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.06430000066757202,"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":0.027300000190734863,"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/inference","display_name":"Inference","score":0.6385999917984009},{"id":"https://openalex.org/keywords/classifier","display_name":"Classifier (UML)","score":0.5734999775886536},{"id":"https://openalex.org/keywords/structured-prediction","display_name":"Structured prediction","score":0.4903999865055084},{"id":"https://openalex.org/keywords/action","display_name":"Action (physics)","score":0.47909998893737793},{"id":"https://openalex.org/keywords/robot","display_name":"Robot","score":0.39559999108314514},{"id":"https://openalex.org/keywords/architecture","display_name":"Architecture","score":0.3799000084400177},{"id":"https://openalex.org/keywords/robotics","display_name":"Robotics","score":0.3379000127315521}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6804999709129333},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6385999917984009},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5855000019073486},{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.5734999775886536},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.4903999865055084},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.47909998893737793},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.4154999852180481},{"id":"https://openalex.org/C90509273","wikidata":"https://www.wikidata.org/wiki/Q11012","display_name":"Robot","level":2,"score":0.39559999108314514},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.3799000084400177},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3605000078678131},{"id":"https://openalex.org/C34413123","wikidata":"https://www.wikidata.org/wiki/Q170978","display_name":"Robotics","level":3,"score":0.3379000127315521},{"id":"https://openalex.org/C146357865","wikidata":"https://www.wikidata.org/wiki/Q1123245","display_name":"Stage (stratigraphy)","level":2,"score":0.33559998869895935},{"id":"https://openalex.org/C4554734","wikidata":"https://www.wikidata.org/wiki/Q593744","display_name":"Knowledge base","level":2,"score":0.31439998745918274},{"id":"https://openalex.org/C2779231336","wikidata":"https://www.wikidata.org/wiki/Q7534724","display_name":"Sketch","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.2712000012397766},{"id":"https://openalex.org/C161759796","wikidata":"https://www.wikidata.org/wiki/Q3982902","display_name":"Teleoperation","level":3,"score":0.25870001316070557},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2572000026702881},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.2540999948978424}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.26801","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.26801","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.26801","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.26801","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"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":{"Vision-Language-Action":[0],"(VLA)":[1],"models":[2],"have":[3],"shown":[4],"strong":[5],"potential":[6],"for":[7,186],"generalizable":[8],"robotic":[9],"manipulation.":[10,192],"During":[11],"fine-tuning,":[12],"however,":[13],"action":[14,91],"supervision":[15,23,58,179],"applies":[16],"equally":[17],"across":[18],"all":[19],"timesteps,":[20],"without":[21,70],"structured":[22,178],"on":[24,123,148],"which":[25],"manipulation":[26,80,172],"stage":[27,74],"the":[28,34,78,88,93,106,113,167],"robot":[29],"is":[30,180],"in":[31,190],"or":[32],"what":[33],"next":[35,94],"gripper-event":[36,47,154],"target":[37,89],"should":[38],"be.":[39],"This":[40],"causes":[41],"failures":[42],"to":[43],"concentrate":[44],"around":[45],"challenging":[46],"transitions.":[48,155],"To":[49],"address":[50],"this,":[51],"we":[52],"propose":[53],"StaKe,":[54],"a":[55,73,83],"plug-in":[56],"auxiliary":[57,102],"framework":[59],"that":[60,76,86,104,132,151,166,177],"automatically":[61],"derives":[62],"two":[63],"complementary":[64],"signals":[65],"from":[66],"demonstration":[67],"gripper":[68,95],"states":[69],"manual":[71],"annotation:":[72],"classifier":[75],"identifies":[77],"current":[79],"stage,":[81],"and":[82,118,126,142,162,183],"keyframe":[84],"predictor":[85],"estimates":[87],"joint":[90],"at":[92],"transition.":[96],"Both":[97],"are":[98],"modeled":[99],"as":[100],"lightweight":[101],"heads":[103],"enrich":[105],"learned":[107,168],"representations":[108,169],"during":[109],"training,":[110],"while":[111],"leaving":[112],"base":[114],"VLA":[115,188],"policy":[116],"architecture":[117],"inference":[119],"loop":[120],"unchanged.":[121],"Experiments":[122],"bimanual":[124],"simulation":[125],"single-arm":[127],"Franka":[128],"real-robot":[129],"tasks":[130,150],"show":[131],"StaKe":[133],"consistently":[134],"improves":[135],"success":[136],"rates":[137],"(relative":[138],"gains":[139],"of":[140],"14%":[141],"56%,":[143],"respectively),":[144],"with":[145],"larger":[146],"improvements":[147],"longer-horizon":[149],"involve":[152],"more":[153],"Ablation":[156],"studies":[157],"validate":[158],"each":[159],"design":[160],"choice,":[161],"qualitative":[163],"analysis":[164],"confirms":[165],"faithfully":[170],"track":[171],"stages.":[173],"These":[174],"results":[175],"indicate":[176],"an":[181],"effective":[182],"general":[184],"strategy":[185],"enhancing":[187],"fine-tuning":[189],"long-horizon":[191],"Project":[193],"website:":[194],"https://hi-yuanxu.github.io/StaKe-Web/":[195]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-27T00:00:00"}