{"id":"https://openalex.org/W7163669142","doi":"https://doi.org/10.48550/arxiv.2606.05597","title":"AsyncWebRL: Efficient Multi-Step RL for Visual Web Agents","display_name":"AsyncWebRL: Efficient Multi-Step RL for Visual Web Agents","publication_year":2026,"publication_date":"2026-06-04","ids":{"openalex":"https://openalex.org/W7163669142","doi":"https://doi.org/10.48550/arxiv.2606.05597"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.05597","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05597","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.2606.05597","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137971718","display_name":"Hao Bai (555291)","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bai, Hao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137996740","display_name":"Rui Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Rui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5072672530","display_name":"Chenlu Ye","orcid":"https://orcid.org/0000-0002-0243-4449"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ye, Chenlu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076767045","display_name":"Spencer Whitehead","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Whitehead, Spencer","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137982015","display_name":"Aviral Kumar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kumar, Aviral","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137941466","display_name":"Tong Zhang (103827)","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Tong","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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.8704000115394592,"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":0.8704000115394592,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.022199999541044235,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.011800000444054604,"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/asynchronous-communication","display_name":"Asynchronous communication","score":0.7620999813079834},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.6744999885559082},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5638999938964844},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5040000081062317},{"id":"https://openalex.org/keywords/state","display_name":"State (computer science)","score":0.450300008058548},{"id":"https://openalex.org/keywords/constant","display_name":"Constant (computer programming)","score":0.4487999975681305},{"id":"https://openalex.org/keywords/aggregate","display_name":"Aggregate (composite)","score":0.42340001463890076}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8108000159263611},{"id":"https://openalex.org/C151319957","wikidata":"https://www.wikidata.org/wiki/Q752739","display_name":"Asynchronous communication","level":2,"score":0.7620999813079834},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.6744999885559082},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5638999938964844},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5040000081062317},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.450300008058548},{"id":"https://openalex.org/C2777027219","wikidata":"https://www.wikidata.org/wiki/Q1284190","display_name":"Constant (computer programming)","level":2,"score":0.4487999975681305},{"id":"https://openalex.org/C4679612","wikidata":"https://www.wikidata.org/wiki/Q866298","display_name":"Aggregate (composite)","level":2,"score":0.42340001463890076},{"id":"https://openalex.org/C171078966","wikidata":"https://www.wikidata.org/wiki/Q111029","display_name":"Root (linguistics)","level":2,"score":0.35359999537467957},{"id":"https://openalex.org/C84945661","wikidata":"https://www.wikidata.org/wiki/Q7366567","display_name":"Root cause","level":2,"score":0.35260000824928284},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.3203999996185303},{"id":"https://openalex.org/C16320812","wikidata":"https://www.wikidata.org/wiki/Q1812200","display_name":"Idle","level":2,"score":0.3100999891757965},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3043999969959259},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3009999990463257},{"id":"https://openalex.org/C35578498","wikidata":"https://www.wikidata.org/wiki/Q193424","display_name":"Web service","level":2,"score":0.2992999851703644},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.296999990940094},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.2922999858856201},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.2578999996185303}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.05597","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05597","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.2606.05597","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05597","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":{"Training":[0],"vision-language":[1],"web":[2],"agents":[3],"with":[4,9,53,131,168],"multi-step":[5,95],"RL":[6],"is":[7],"compute-intensive,":[8],"two":[10,54],"dominant":[11],"forms":[12],"of":[13,101,152],"inefficiency:":[14,105],"idle":[15],"GPUs":[16],"in":[17,94],"synchronous":[18,81],"RL,":[19],"and":[20,26,47,62,103],"trajectories":[21,139],"that":[22,66],"use":[23],"more":[24],"steps":[25],"tokens":[27],"than":[28,111],"necessary.":[29],"We":[30],"present":[31],"AsyncWebRL,":[32],"which":[33],"addresses":[34],"both.":[35],"On":[36,84],"the":[37,77,85,90,98,115,122,153,156,164,169,173],"system":[38],"side,":[39,87],"an":[40,58],"asynchronous":[41],"design":[42],"overlaps":[43],"rollout,":[44],"gradient":[45,117],"update,":[46],"policy":[48,123],"refresh":[49],"across":[50],"iterations,":[51],"paired":[52],"web-agent-specific":[55],"adaptations,":[56],"namely":[57],"everlasting":[59],"rollout":[60],"pool":[61],"lightweight":[63],"screenshot":[64],"handling,":[65],"together":[67],"deliver":[68],"up":[69],"to":[70],"a":[71,132,148],"$2.9\\times$":[72],"end-to-end":[73],"training-throughput":[74],"speedup":[75],"over":[76,163],"previously":[78],"fastest":[79],"open":[80],"pipeline":[82],"(WebGym).":[83],"algorithmic":[86],"we":[88],"identify":[89],"per-trajectory":[91],"normalizer":[92],"$1/|\u03c4_i|$":[93,130],"GRPO":[96],"as":[97],"root":[99],"cause":[100],"trajectory-level":[102],"token-level":[104],"because":[106],"failures":[107],"are":[108],"systematically":[109],"longer":[110],"successes,":[112],"it":[113],"down-weights":[114],"negative":[116],"on":[118,155,172,178,182],"failed":[119],"tokens,":[120],"so":[121],"keeps":[124],"producing":[125],"verbose":[126],"memory":[127],"schemas.":[128],"Replacing":[129],"constant":[133],"$1/k$":[134],"breaks":[135],"this":[136],"coupling,":[137],"contracting":[138],"while":[140],"preserving":[141],"aggregate":[142],"success.":[143],"Together,":[144],"these":[145],"contributions":[146],"set":[147],"new":[149],"open-source":[150],"state":[151],"art":[154],"WebGym":[157],"out-of-distribution":[158],"test":[159],"split":[160],"(+5.8%":[161],"relative":[162,177,181],"42.9%":[165],"prior":[166],"best),":[167],"largest":[170],"gains":[171],"harder":[174],"slices":[175],"(+42%":[176],"Medium,":[179],"+48%":[180],"Hard).":[183]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-06T00:00:00"}