{"id":"https://openalex.org/W7166879155","doi":"https://doi.org/10.1613/jair.1.20097","title":"Language Model Self-improvement by Reinforcement Learning Contemplation without External Supervision","display_name":"Language Model Self-improvement by Reinforcement Learning Contemplation without External Supervision","publication_year":2026,"publication_date":"2026-07-01","ids":{"openalex":"https://openalex.org/W7166879155","doi":"https://doi.org/10.1613/jair.1.20097"},"language":null,"primary_location":{"id":"doi:10.1613/jair.1.20097","is_oa":true,"landing_page_url":"https://doi.org/10.1613/jair.1.20097","pdf_url":"https://www.jair.org/index.php/jair/article/download/20097/27317","source":{"id":"https://openalex.org/S139930977","display_name":"Journal of Artificial Intelligence Research","issn_l":"1076-9757","issn":["1076-9757","1943-5037"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310315760","host_organization_name":"AI Access Foundation","host_organization_lineage":["https://openalex.org/P4310315760"],"host_organization_lineage_names":["AI Access Foundation"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Journal of Artificial Intelligence Research","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://www.jair.org/index.php/jair/article/download/20097/27317","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5008843087","display_name":"Jing-Cheng Pang","orcid":"https://orcid.org/0000-0002-6086-6979"},"institutions":[{"id":"https://openalex.org/I36399199","display_name":"Nanjing University of Science and Technology","ror":"https://ror.org/00xp9wg62","country_code":"CN","type":"education","lineage":["https://openalex.org/I36399199"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jing-Cheng Pang","raw_affiliation_strings":["a:1:{s:5:\"en_US\";s:18:\"Nanjing University\";}"],"raw_orcid":"https://orcid.org/0000-0002-6086-6979","affiliations":[{"raw_affiliation_string":"a:1:{s:5:\"en_US\";s:18:\"Nanjing University\";}","institution_ids":["https://openalex.org/I36399199"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139852729","display_name":"Kaiyuan Li","orcid":"https://orcid.org/0009-0009-1331-9604"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kaiyuan Li","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0009-0009-1331-9604","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139730779","display_name":"Pengyuan Wang","orcid":"https://orcid.org/0009-0006-1545-9457"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pengyuan Wang","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0009-0006-1545-9457","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5019686106","display_name":"Xiong-Hui Chen","orcid":"https://orcid.org/0000-0003-4911-3148"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiong-Hui Chen","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0003-4911-3148","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100622267","display_name":"Jiacheng Xu","orcid":"https://orcid.org/0009-0008-4315-6976"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiacheng Xu","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0009-0008-4315-6976","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035138959","display_name":"Zongzhang Zhang","orcid":"https://orcid.org/0000-0002-9238-4747"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zongzhang Zhang","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0002-9238-4747","affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100342263","display_name":"Yang Yu","orcid":"https://orcid.org/0000-0002-1732-9545"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang Yu","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0002-1732-9545","affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.88496908,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"86","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":0.5745000243186951,"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":0.5745000243186951,"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":0.03660000115633011,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.03539999946951866,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.786899983882904},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.7617999911308289},{"id":"https://openalex.org/keywords/automatic-summarization","display_name":"Automatic summarization","score":0.6933000087738037},{"id":"https://openalex.org/keywords/heuristic","display_name":"Heuristic","score":0.5443000197410583},{"id":"https://openalex.org/keywords/contemplation","display_name":"Contemplation","score":0.4530999958515167},{"id":"https://openalex.org/keywords/reinforcement","display_name":"Reinforcement","score":0.3986000120639801},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.3817000091075897},{"id":"https://openalex.org/keywords/question-answering","display_name":"Question answering","score":0.33550000190734863}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.786899983882904},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.7617999911308289},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7319999933242798},{"id":"https://openalex.org/C170858558","wikidata":"https://www.wikidata.org/wiki/Q1394144","display_name":"Automatic summarization","level":2,"score":0.6933000087738037},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5669999718666077},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.5443000197410583},{"id":"https://openalex.org/C178288346","wikidata":"https://www.wikidata.org/wiki/Q1123855","display_name":"Contemplation","level":2,"score":0.4530999958515167},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.3986000120639801},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.39089998602867126},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3817000091075897},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.37529999017715454},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.33550000190734863},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.32019999623298645},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.31119999289512634},{"id":"https://openalex.org/C74672266","wikidata":"https://www.wikidata.org/wiki/Q815859","display_name":"Language acquisition","level":2,"score":0.30570000410079956},{"id":"https://openalex.org/C2779313563","wikidata":"https://www.wikidata.org/wiki/Q17072565","display_name":"On Language","level":2,"score":0.28760001063346863},{"id":"https://openalex.org/C94922259","wikidata":"https://www.wikidata.org/wiki/Q33215","display_name":"Constructed language","level":2,"score":0.2808000147342682},{"id":"https://openalex.org/C129792486","wikidata":"https://www.wikidata.org/wiki/Q1050419","display_name":"Language identification","level":3,"score":0.2793000042438507},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.27309998869895935},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.26440000534057617},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.2574000060558319},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.2500999867916107}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1613/jair.1.20097","is_oa":true,"landing_page_url":"https://doi.org/10.1613/jair.1.20097","pdf_url":"https://www.jair.org/index.php/jair/article/download/20097/27317","source":{"id":"https://openalex.org/S139930977","display_name":"Journal of Artificial Intelligence Research","issn_l":"1076-9757","issn":["1076-9757","1943-5037"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310315760","host_organization_name":"AI Access Foundation","host_organization_lineage":["https://openalex.org/P4310315760"],"host_organization_lineage_names":["AI Access Foundation"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Journal of Artificial Intelligence Research","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1613/jair.1.20097","is_oa":true,"landing_page_url":"https://doi.org/10.1613/jair.1.20097","pdf_url":"https://www.jair.org/index.php/jair/article/download/20097/27317","source":{"id":"https://openalex.org/S139930977","display_name":"Journal of Artificial Intelligence Research","issn_l":"1076-9757","issn":["1076-9757","1943-5037"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310315760","host_organization_name":"AI Access Foundation","host_organization_lineage":["https://openalex.org/P4310315760"],"host_organization_lineage_names":["AI Access Foundation"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Journal of Artificial Intelligence Research","raw_type":"journal-article"},"sustainable_development_goals":[{"score":0.7400563359260559,"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education"}],"awards":[{"id":"https://openalex.org/G6712651544","display_name":null,"funder_award_id":"B26023","funder_id":"https://openalex.org/F4320336698","funder_display_name":"Overseas Expertise Introduction Center for Discipline Innovation of Food Nutrition and Human Health (111 Center)"}],"funders":[{"id":"https://openalex.org/F4320321106","display_name":"Ministry of Education of the People's Republic of China","ror":"https://ror.org/01mv9t934"},{"id":"https://openalex.org/F4320336698","display_name":"Overseas Expertise Introduction Center for Discipline Innovation of Food Nutrition and Human Health (111 Center)","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166879155.pdf","grobid_xml":"https://content.openalex.org/works/W7166879155.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Language":[0],"model":[1,33,60,71,129,197],"self-improvement":[2],"(LMSI)":[3],"techniques":[4],"have":[5],"recently":[6],"gained":[7],"significant":[8],"attention":[9],"as":[10],"they":[11],"improve":[12],"language":[13,48,70,96,128,220],"models":[14,97,109,189],"without":[15,235],"requiring":[16],"external":[17,236],"supervision.":[18,237],"A":[19],"notable":[20],"approach":[21],"is":[22,61,93],"reinforcement":[23,41],"learning":[24,42],"from":[25],"AI":[26,36,59],"feedback":[27],"(RLAIF),":[28],"which":[29,67],"trains":[30],"a":[31,54,81,143,159,173,231],"reward":[32],"based":[34],"on":[35,53,142,213,226],"preference":[37],"data":[38],"and":[39,119,126,154,172,196,201,223],"employs":[40],"(RL)":[43],"algorithm":[44],"to":[45,63,72,98,103,132,166,188,194],"train":[46],"the":[47,58,69,100,114,117,123,138,162],"model.":[49],"However,":[50],"RLAIF":[51],"relies":[52],"heuristic":[55],"assumption":[56],"that":[57,91,210],"able":[62],"provide":[64],"effective":[65],"feedback,":[66],"requires":[68],"possess":[73],"solid":[74],"capability.":[75],"In":[76,182],"this":[77],"paper,":[78],"we":[79],"present":[80],"novel":[82],"LMSI":[83],"method,":[84],"Reinforcement":[85],"Learning":[86],"Contemplation":[87],"(RLC).":[88],"We":[89,136,207],"disclose":[90],"it":[92],"simpler":[94],"for":[95,107,168,177],"evaluate":[99],"text":[101],"than":[102],"generate":[104],"it,":[105],"even":[106],"small":[108],"under":[110],"1B":[111],"parameters.":[112],"Leveraging":[113],"gap":[115],"between":[116],"evaluation":[118,222],"generation,":[120],"RLC":[121,141,184,218],"evaluates":[122],"generated":[124],"answers":[125],"updates":[127],"using":[130],"RL":[131],"maximize":[133],"self-evaluation":[134],"scores.":[135],"demonstrate":[137],"effectiveness":[139],"of":[140,146,190,216],"wide":[144],"range":[145],"challenging":[147],"tasks,":[148,171],"including":[149],"reasoning,":[150],"summarization,":[151],"conditioned":[152],"generation":[153,224],"emotion":[155],"recognition,":[156],"resulting":[157],"in":[158,161,175],"increase":[160],"answering":[163],"accuracy":[164],"(31.23%":[165],"37.09%)":[167],"BigBench-hard":[169],"reasoning":[170],"rise":[174],"BERTScore":[176],"CNN/Daily":[178],"Mail":[179],"summarization":[180],"tasks.":[181],"addition,":[183],"can":[185],"be":[186],"applied":[187],"different":[191],"sizes":[192],"(80M":[193],"3B)":[195],"structures":[198],"(FLAN-T5,":[199],"LLAMA-3.2":[200],"QWEN-2.5),":[202],"showcasing":[203],"its":[204],"broad":[205],"applicability.":[206],"further":[208],"verify":[209],"when":[211],"training":[212],"larger":[214],"scale":[215],"dataset,":[217],"improves":[219],"model\u2019s":[221],"ability":[225],"unseen":[227],"tasks":[228],"simultaneously,":[229],"enabling":[230],"general":[232],"capability":[233],"improvement":[234]},"counts_by_year":[],"updated_date":"2026-07-03T06:15:21.484131","created_date":"2026-07-02T00:00:00"}