{"id":"https://openalex.org/W7213675698","doi":null,"title":"Reflective Recovery: A Self-Supervised Method for Reasoning by Learning from Mistakes","display_name":"Reflective Recovery: A Self-Supervised Method for Reasoning by Learning from Mistakes","publication_year":2026,"publication_date":"2026-07-24","ids":{"openalex":"https://openalex.org/W7213675698"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.19156","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.19156","pdf_url":"https://arxiv.org/pdf/2609.19156","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"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.19156","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5152178645","display_name":"Qirui Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I76130692","display_name":"Zhejiang University","ror":"https://ror.org/00a2xv884","country_code":"CN","type":"education","lineage":["https://openalex.org/I76130692"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen, Qirui","raw_affiliation_strings":["15958 , Vienna , Austria","Zhejiang University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"15958 , Vienna , Austria","institution_ids":[]},{"raw_affiliation_string":"Zhejiang University","institution_ids":["https://openalex.org/I76130692"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5028182071","display_name":"Renjie Pi","orcid":"https://orcid.org/0000-0002-5033-7096","observed_orcids":["https://orcid.org/0000-0002-5033-7096"]},"institutions":[{"id":"https://openalex.org/I200769079","display_name":"Hong Kong University of Science and Technology","ror":"https://ror.org/00q4vv597","country_code":"HK","type":"education","lineage":["https://openalex.org/I200769079"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Pi, Renjie","raw_affiliation_strings":["The Hong Kong University of Science and Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology","institution_ids":["https://openalex.org/I200769079"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101843320","display_name":"Jiahui Gao","orcid":"https://orcid.org/0000-0001-7243-4981","observed_orcids":["https://orcid.org/0000-0001-7243-4981"]},"institutions":[{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Gao, Jiahui","raw_affiliation_strings":["The University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Hong Kong","institution_ids":["https://openalex.org/I889458895"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5014554970","display_name":"Lingpeng Kong","orcid":"https://orcid.org/0000-0002-9033-2724","observed_orcids":["https://orcid.org/0000-0002-9033-2724"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kong, Lingpeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"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/T10028","display_name":"Topic Modeling","score":0.7778000235557556,"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":0.7778000235557556,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.054499998688697815,"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":0.026499999687075615,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/self-supervised-learning","display_name":"self-supervised learning","score":0.7089999914169312},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.8140000104904175},{"id":"https://openalex.org/keywords/reasoning","display_name":"reasoning","score":0.453000009059906},{"id":"https://openalex.org/keywords/self-correction","display_name":"self-correction","score":0.7099999785423279},{"id":"https://openalex.org/keywords/scaling-collapse","display_name":"scaling collapse","score":0.8500000238418579},{"id":"https://openalex.org/keywords/imitation-learning","display_name":"imitation learning","score":0.8539999723434448},{"id":"https://openalex.org/keywords/fine-tuning","display_name":"fine-tuning","score":0.5540000200271606},{"id":"https://openalex.org/keywords/reflective-thinking","display_name":"reflective thinking","score":0.47099998593330383},{"id":"https://openalex.org/keywords/mathematical-problem-solving","display_name":"mathematical problem solving","score":0.4169999957084656}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6668999791145325},{"id":"https://openalex.org/C30038468","wikidata":"https://www.wikidata.org/wiki/Q4354775","display_name":"Memorization","level":2,"score":0.6477000117301941},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.6424000263214111},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.6384999752044678},{"id":"https://openalex.org/C2776372474","wikidata":"https://www.wikidata.org/wiki/Q508291","display_name":"Simplicity","level":2,"score":0.5569000244140625},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5239999890327454},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.43160000443458557},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.3765000104904175},{"id":"https://openalex.org/C126388530","wikidata":"https://www.wikidata.org/wiki/Q1131737","display_name":"Imitation","level":2,"score":0.34599998593330383},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.3077000081539154},{"id":"https://openalex.org/C65682993","wikidata":"https://www.wikidata.org/wiki/Q1056451","display_name":"Reflection (computer programming)","level":2,"score":0.2969000041484833},{"id":"https://openalex.org/C2777617010","wikidata":"https://www.wikidata.org/wiki/Q18957","display_name":"Mainstream","level":2,"score":0.2809999883174896},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2736999988555908},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.2637999951839447},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.26100000739097595},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.25519999861717224}],"mesh":[],"locations_count":1,"locations":[{"id":"pmh:oai:arXiv.org:2609.19156","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.19156","pdf_url":"https://arxiv.org/pdf/2609.19156","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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.19156","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.19156","pdf_url":"https://arxiv.org/pdf/2609.19156","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":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7213675698.pdf","grobid_xml":"https://content.openalex.org/works/W7213675698.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7166802559","https://openalex.org/W7160616492","https://openalex.org/W7117857712","https://openalex.org/W7117174392","https://openalex.org/W4388185044","https://openalex.org/W4402672106","https://openalex.org/W7125100858","https://openalex.org/W7139910495","https://openalex.org/W7221550777","https://openalex.org/W7140225061"],"abstract_inverted_index":{"Data-driven":[0],"fine-tuning":[1],"is":[2,40,62,65],"widely":[3],"adopted":[4],"to":[5,14,46,68,78,128,143,150,186,194,212,224],"enhance":[6],"reasoning":[7,29,107],"in":[8],"Large":[9],"Language":[10],"Models":[11],"(LLMs)":[12],"due":[13],"its":[15],"simplicity":[16],"and":[17,64,80,125,152,191,209],"efficiency.":[18],"However,":[19,50],"mainstream":[20],"imitation":[21],"learning":[22],"methods":[23],"that":[24,58,104,202],"rely":[25],"exclusively":[26],"on":[27,164,171,188,196],"perfect":[28],"trajectories":[30,140],"suffer":[31],"from":[32,138,159,184,192,221],"a":[33,98,218],"Scaling":[34],"Collapse:":[35],"when":[36],"the":[37,74,86,130,205],"problem":[38],"set":[39],"limited,":[41],"increasing":[42],"positive":[43],"examples":[44],"fails":[45],"yield":[47],"continuous":[48],"improvement.":[49],"during":[51,155],"inference,":[52],"an":[53],"LLM":[54,75,131],"can":[55],"not":[56],"guarantee":[57],"every":[59],"intermediate":[60],"step":[61],"correct":[63,153],"therefore":[66],"prone":[67],"errors.":[69],"Once":[70],"such":[71],"errors":[72],"arise,":[73],"often":[76],"struggles":[77],"recover":[79],"may":[81],"be":[82],"further":[83],"misled":[84],"by":[85],"accumulation":[87],"of":[88,118],"previous":[89],"mistakes.":[90],"To":[91],"address":[92],"this,":[93],"we":[94,114],"propose":[95],"Reflective":[96,174],"Recovery,":[97],"simple":[99],"yet":[100],"effective":[101],"self-supervised":[102],"approach":[103],"transforms":[105],"failed":[106,119,139],"attempts":[108],"into":[109],"recovery":[110,158],"training":[111],"data.":[112],"Specifically,":[113],"extract":[115],"initial":[116],"segments":[117,137],"trajectories,":[120],"concatenate":[121],"them":[122,127],"with":[123],"prompts,":[124],"use":[126],"guide":[129],"toward":[132],"valid":[133],"solutions.":[134],"Because":[135],"these":[136],"are":[141],"likely":[142],"contain":[144],"errors,":[145],"this":[146],"process":[147],"teaches":[148],"models":[149,211],"recognize":[151],"mistakes":[154],"reasoning,":[156],"enabling":[157],"erroneous":[160],"states":[161],"without":[162],"relying":[163],"external":[165],"critics":[166],"or":[167],"reward":[168],"models.":[169],"Evaluated":[170],"extensive":[172],"benchmarks,":[173],"Recovery":[175],"significantly":[176],"improves":[177],"performance.":[178],"On":[179],"DeepSeek-R1-Distill-Qwen-7B,":[180],"it":[181,203],"boosts":[182],"accuracy":[183],"30.0%":[185],"37.5%":[187],"AIME":[189],"2025":[190],"37.6%":[193],"47.8%":[195],"Minerva.":[197],"More":[198],"importantly,":[199],"analyses":[200],"demonstrate":[201],"breaks":[204],"scaling":[206],"collapse":[207],"barrier":[208],"enables":[210],"develop":[213],"emergent":[214],"self-correction":[215],"behaviors,":[216],"representing":[217],"paradigm":[219],"shift":[220],"outcome-oriented":[222],"memorization":[223],"process-oriented":[225],"reflective":[226],"reasoning.":[227]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-20T00:00:00"}