{"id":"https://openalex.org/W7161061303","doi":"https://doi.org/10.48550/arxiv.2605.11235","title":"Internalizing Curriculum Judgment for LLM Reinforcement Fine-Tuning","display_name":"Internalizing Curriculum Judgment for LLM Reinforcement Fine-Tuning","publication_year":2026,"publication_date":"2026-05-11","ids":{"openalex":"https://openalex.org/W7161061303","doi":"https://doi.org/10.48550/arxiv.2605.11235"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.11235","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11235","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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.2605.11235","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136063866","display_name":"Han Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Han","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136005622","display_name":"Yining Ma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Yining","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136078094","display_name":"Karthick Gunasekaran","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gunasekaran, Karthick","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101689740","display_name":"Bharathan Balaji","orcid":"https://orcid.org/0000-0002-9490-2018"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Balaji, Bharathan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136081314","display_name":"Zheng Du","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Du, Zheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136059603","display_name":"Shiv Vitaladevuni","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Vitaladevuni, Shiv","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136016988","display_name":"Cathy Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Cathy","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":7,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.3237999975681305,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.3237999975681305,"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/T10028","display_name":"Topic Modeling","score":0.10750000178813934,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.060100000351667404,"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/metis","display_name":"Metis","score":0.8073999881744385},{"id":"https://openalex.org/keywords/heuristics","display_name":"Heuristics","score":0.758899986743927},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.6901000142097473},{"id":"https://openalex.org/keywords/curriculum","display_name":"Curriculum","score":0.6388000249862671},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.4970000088214874},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.48890000581741333},{"id":"https://openalex.org/keywords/reinforcement","display_name":"Reinforcement","score":0.3813999891281128}],"concepts":[{"id":"https://openalex.org/C2780705272","wikidata":"https://www.wikidata.org/wiki/Q19903554","display_name":"Metis","level":2,"score":0.8073999881744385},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.758899986743927},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6901000142097473},{"id":"https://openalex.org/C47177190","wikidata":"https://www.wikidata.org/wiki/Q207137","display_name":"Curriculum","level":2,"score":0.6388000249862671},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6046000123023987},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.4970000088214874},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.48890000581741333},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.39149999618530273},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.3813999891281128},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.3138999938964844},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.30970001220703125},{"id":"https://openalex.org/C132758656","wikidata":"https://www.wikidata.org/wiki/Q5307365","display_name":"Dreyfus model of skill acquisition","level":2,"score":0.30649998784065247},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.3034999966621399},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.28850001096725464},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.28110000491142273},{"id":"https://openalex.org/C130440534","wikidata":"https://www.wikidata.org/wiki/Q14946528","display_name":"Conflation","level":2,"score":0.2741999924182892},{"id":"https://openalex.org/C511192102","wikidata":"https://www.wikidata.org/wiki/Q5156948","display_name":"Comprehension","level":2,"score":0.27399998903274536},{"id":"https://openalex.org/C2780586970","wikidata":"https://www.wikidata.org/wiki/Q1357284","display_name":"Popularity","level":2,"score":0.2605000138282776}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.11235","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11235","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.11235","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11235","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.7737164497375488}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"In":[0,31],"LLM":[1,168],"Reinforcement":[2],"Fine-Tuning":[3],"(RFT),":[4],"curriculum":[5,16,45,164],"learning":[6,75],"drives":[7],"both":[8],"efficiency":[9],"and":[10,93,102,124,152,161],"performance.":[11],"Yet,":[12],"current":[13],"methods":[14],"externalize":[15],"judgment":[17,46,92],"via":[18],"handcrafted":[19,150],"heuristics":[20,151],"or":[21],"auxiliary":[22,153],"models,":[23,154],"risking":[24],"misalignment":[25],"with":[26],"the":[27,83,89,98,108],"policy's":[28],"training":[29,70,84],"dynamics.":[30],"this":[32,65],"paper,":[33],"we":[34],"introduce":[35],"METIS":[36,63,87,136],"(METacognitive":[37],"Internalized":[38],"Self-judgment),":[39],"a":[40,48,52,103,117,158],"novel":[41],"framework":[42],"that":[43,55],"internalizes":[44],"as":[47,72,116],"native":[49],"capability.":[50],"Leveraging":[51],"critical":[53],"observation":[54],"within-prompt":[56],"reward":[57],"variance":[58],"effectively":[59],"gauges":[60],"prompt":[61],"informativeness,":[62],"predicts":[64],"metric":[66],"based":[67],"on":[68],"recent":[69],"outcomes":[71],"lightweight":[73],"in-context":[74],"examples.":[76],"This":[77,106],"intrinsic":[78],"self-judgment":[79,104],"then":[80],"dynamically":[81],"dictates":[82],"allocation.":[85],"Moreover,":[86],"closes":[88],"loop":[90],"between":[91],"optimization":[94],"by":[95,144],"jointly":[96],"optimizing":[97],"standard":[99],"RFT":[100,126],"rewards":[101],"reward.":[105],"allows":[107],"policy":[109],"to":[110,113,133,146],"learn":[111,114],"what":[112],"next,":[115],"form":[118],"of":[119],"metacognition.":[120],"Across":[121],"extensive":[122],"discrete":[123],"continuous":[125],"benchmarks":[127],"from":[128],"mathematical":[129],"reasoning,":[130],"code":[131],"generation,":[132],"agentic":[134],"function-calling,":[135],"consistently":[137],"delivers":[138],"superior":[139],"performance":[140],"while":[141],"accelerating":[142],"convergence":[143],"up":[145],"67%.":[147],"By":[148],"bypassing":[149],"our":[155],"work":[156],"establishes":[157],"simple,":[159],"closed-loop,":[160],"highly":[162],"efficient":[163],"internalization":[165],"paradigm":[166],"for":[167],"reinforcement":[169],"fine-tuning.":[170]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-05-14T00:00:00"}