{"id":"https://openalex.org/W7163001932","doi":"https://doi.org/10.48550/arxiv.2605.31228","title":"EchoRL: Reinforcement Learning via Rollout Echoing","display_name":"EchoRL: Reinforcement Learning via Rollout Echoing","publication_year":2026,"publication_date":"2026-05-29","ids":{"openalex":"https://openalex.org/W7163001932","doi":"https://doi.org/10.48550/arxiv.2605.31228"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.31228","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.31228","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.31228","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137539939","display_name":"Jinhe Bi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bi, Jinhe","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137555437","display_name":"Aniri","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Aniri","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137563993","display_name":"Minglai Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Minglai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137513391","display_name":"Xingcheng Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Xingcheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137608566","display_name":"Wenke Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Wenke","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132755543","display_name":"Sikuan Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yan, Sikuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137545501","display_name":"Yujun Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yujun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101184662","display_name":"Zixuan Cao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cao, Zixuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137567535","display_name":"Michael F\u00e4rber","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"F\u00e4rber, Michael","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137537244","display_name":"Xun Xiao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiao, Xun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137545977","display_name":"Volker Tresp","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tresp, Volker","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137520608","display_name":"Yunpu Ma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Yunpu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":12,"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/T10028","display_name":"Topic Modeling","score":0.35109999775886536,"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/T10028","display_name":"Topic Modeling","score":0.35109999775886536,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.09790000319480896,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.0617000013589859,"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":0.666100025177002},{"id":"https://openalex.org/keywords/entropy","display_name":"Entropy (arrow of time)","score":0.399399995803833},{"id":"https://openalex.org/keywords/reinforcement","display_name":"Reinforcement","score":0.38449999690055847},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.30169999599456787},{"id":"https://openalex.org/keywords/fraction","display_name":"Fraction (chemistry)","score":0.2842999994754791}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6723999977111816},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.666100025177002},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.399399995803833},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.38449999690055847},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.35010001063346863},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.30169999599456787},{"id":"https://openalex.org/C149629883","wikidata":"https://www.wikidata.org/wiki/Q660926","display_name":"Fraction (chemistry)","level":2,"score":0.2842999994754791},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2606000006198883},{"id":"https://openalex.org/C85847156","wikidata":"https://www.wikidata.org/wiki/Q59015987","display_name":"Verifiable secret sharing","level":3,"score":0.25920000672340393},{"id":"https://openalex.org/C132758656","wikidata":"https://www.wikidata.org/wiki/Q5307365","display_name":"Dreyfus model of skill acquisition","level":2,"score":0.25119999051094055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.31228","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.31228","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.31228","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.31228","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":"article"},"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":{"Reinforcement":[0],"Learning":[1],"with":[2,101,193],"Verifiable":[3],"Rewards":[4],"is":[5,139],"an":[6,146,164],"effective":[7],"route":[8],"for":[9,77,126],"post-training":[10,184,192],"to":[11,132],"strengthen":[12],"the":[13,24,31,48,54,75,83,102,112,129,135,169],"reasoning":[14],"capability":[15],"of":[16,42,90],"large":[17],"language":[18],"models.":[19],"However,":[20],"as":[21,69,163],"training":[22,32,84,136],"proceeds,":[23],"learning":[25,96],"signal":[26,167],"can":[27],"collapse":[28],"thus":[29],"makes":[30],"gain":[33],"become":[34,45],"marginal":[35],"and":[36,157,180],"ineffective.":[37],"Specifically,":[38],"a":[39,140],"growing":[40],"fraction":[41],"prompts'":[43],"rollouts":[44,50,92,131,150],"advantage-degenerated:":[46],"all":[47],"self-generated":[49],"show":[51],"verified-success,":[52],"making":[53],"standard":[55],"deviation":[56],"over":[57],"their":[58,153],"rewards":[59],"be":[60],"zero;":[61],"accordingly":[62],"each":[63],"rollout's":[64],"advantage":[65],"becomes":[66],"degenerated":[67],"(zero)":[68],"well.":[70],"Given":[71],"such":[72],"rollouts'":[73],"advantages,":[74],"policy-gradient":[76],"model":[78],"optimization":[79],"eventually":[80],"vanishes,":[81],"capping":[82],"performance.":[85,137],"We":[86],"argue":[87],"that":[88,143,187],"some":[89],"these":[91],"still":[93],"contain":[94],"valuable":[95],"signals":[97],"but":[98],"unfortunately":[99],"omitted":[100],"existing":[103],"RLVR":[104,183,191],"methods.":[105],"In":[106],"this":[107,160],"paper,":[108],"inspired":[109],"through":[110],"analyzing":[111],"entropy":[113,155],"pattern":[114],"behind":[115],"golden":[116],"trajectories":[117],"produced":[118],"by":[119],"external":[120],"expert":[121],"models,":[122],"we":[123],"propose":[124],"EchoRL":[125,138,188],"better":[127],"exploiting":[128],"advantage-degenerated":[130],"further":[133],"improve":[134],"lightweight":[141],"module":[142],"first":[144],"identifies":[145],"EchoClip":[147],"from":[148],"verified-success":[149],"based":[151],"on":[152],"step-level":[154],"values,":[156],"then":[158],"feeds":[159],"clip":[161],"back":[162],"auxiliary":[165],"supervision":[166],"in":[168],"RL":[170],"objective.":[171],"Extensive":[172],"experiments":[173],"across":[174],"10":[175],"benchmarks,":[176],"5":[177],"LLM":[178],"backbones,":[179],"4":[181],"popular":[182],"methods":[185],"demonstrate":[186],"consistently":[189],"improves":[190],"minimal":[194],"overhead.":[195]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-06-02T00:00:00"}