{"id":"https://openalex.org/W7168348247","doi":"https://doi.org/10.48550/arxiv.2607.12790","title":"Who Grades the Grader? Co-Evolving Evaluation Metrics and Skills for Self-Improving LLM Agents","display_name":"Who Grades the Grader? Co-Evolving Evaluation Metrics and Skills for Self-Improving LLM Agents","publication_year":2026,"publication_date":"2026-07-14","ids":{"openalex":"https://openalex.org/W7168348247","doi":"https://doi.org/10.48550/arxiv.2607.12790"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2607.12790","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.12790","pdf_url":null,"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,"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.2607.12790","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5140844955","display_name":"Xing Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Xing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140754557","display_name":"Guanghui Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Guanghui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140776529","display_name":"Yanwei Cui","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cui, Yanwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140752876","display_name":"Ziyuan Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Ziyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140805981","display_name":"Wei Qiu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qiu, Wei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140859391","display_name":"Bing Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Bing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5140804916","display_name":"Peiyang He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Peiyang","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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.11540000140666962,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.11540000140666962,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.0722000002861023,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.05719999969005585,"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/yardstick","display_name":"Yardstick","score":0.7444000244140625},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.7124999761581421},{"id":"https://openalex.org/keywords/audit","display_name":"Audit","score":0.5},{"id":"https://openalex.org/keywords/metric-system","display_name":"Metric system","score":0.498199999332428},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.36500000953674316},{"id":"https://openalex.org/keywords/detector","display_name":"Detector","score":0.3384000062942505}],"concepts":[{"id":"https://openalex.org/C82261069","wikidata":"https://www.wikidata.org/wiki/Q1925365","display_name":"Yardstick","level":2,"score":0.7444000244140625},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.7124999761581421},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.510200023651123},{"id":"https://openalex.org/C199521495","wikidata":"https://www.wikidata.org/wiki/Q181487","display_name":"Audit","level":2,"score":0.5},{"id":"https://openalex.org/C193429382","wikidata":"https://www.wikidata.org/wiki/Q232405","display_name":"Metric system","level":2,"score":0.498199999332428},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.40939998626708984},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.36500000953674316},{"id":"https://openalex.org/C94915269","wikidata":"https://www.wikidata.org/wiki/Q1834857","display_name":"Detector","level":2,"score":0.3384000062942505},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.33059999346733093},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.3303000032901764},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.3264999985694885},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3190000057220459},{"id":"https://openalex.org/C139002025","wikidata":"https://www.wikidata.org/wiki/Q3001212","display_name":"Lift (data mining)","level":2,"score":0.3160000145435333},{"id":"https://openalex.org/C82214349","wikidata":"https://www.wikidata.org/wiki/Q657339","display_name":"Software metric","level":5,"score":0.27300000190734863},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.26899999380111694},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.2662999927997589},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.26489999890327454},{"id":"https://openalex.org/C146849305","wikidata":"https://www.wikidata.org/wiki/Q370766","display_name":"Ground truth","level":2,"score":0.25540000200271606}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2607.12790","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.12790","pdf_url":null,"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,"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.2607.12790","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.12790","pdf_url":null,"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,"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":[{"score":0.5363229513168335,"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Self-evolving":[0],"agent":[1],"systems":[2],"create,":[3],"revise,":[4],"and":[5,60,85,191,221,236],"retire":[6],"their":[7],"own":[8],"skills,":[9],"but":[10],"every":[11],"such":[12],"loop":[13,38],"assumes":[14],"a":[15,47,55,114,144,174,202,214],"reliable":[16],"evaluation":[17],"metric":[18,30,118,142,180,194,200],"already":[19],"exists.":[20],"In":[21],"many":[22],"real":[23],"applications":[24],"none":[25],"does.":[26],"We":[27],"show":[28],"the":[29,34,70,77,125,137,141,150,156,178,199,209,228],"itself":[31],"can":[32],"be":[33],"evolving":[35],"object:":[36],"our":[37],"searches":[39],"compositions":[40],"of":[41,166,208],"small":[42],"typed":[43],"drawback":[44],"detectors":[45],"under":[46],"full":[48],"evolutionary":[49],"lifecycle,":[50],"selecting":[51],"for":[52],"agreement":[53,108],"with":[54,109,201],"ten-item":[56],"anchored":[57],"reference":[58],"set":[59,116],"regularizing":[61],"by":[62],"consensus":[63],"over":[64],"unlabeled":[65],"outputs.":[66],"What":[67],"evolves":[68],"is":[69,82,89,99,132],"function":[71],"that":[72,117,168],"grades":[73],"one":[74,237],"output,":[75],"never":[76,120],"fixed":[78],"task":[79,170],"sets":[80],"it":[81,105,129,235],"scored":[83],"on,":[84],"what":[86],"comes":[87],"out":[88],"an":[90,95,192,231],"inspectable":[91],"expression":[92],"rather":[93],"than":[94],"opaque":[96],"judge.":[97],"It":[98],"also":[100],"valid:":[101],"on":[102,113],"code":[103],"generation":[104],"gains":[106],"0.21":[107],"hidden":[110],"ground":[111,211],"truth":[112,212],"locked":[115],"selection":[119],"reads":[121],"(paired":[122],"$p=0.014$),":[123],"beating":[124],"bare":[126],"LLM":[127],"judge":[128,233],"contains.":[130],"Validity":[131],"where":[133],"safety":[134],"lives:":[135],"removing":[136,149],"anchor":[138],"guards":[139],"collapses":[140],"into":[143],"vacuous":[145],"always-pass":[146],"detector":[147,151,239],"while":[148],"lifecycle":[152],"does":[153],"not,":[154],"inverting":[155],"lesson":[157],"from":[158],"skill":[159,204],"evolution.":[160],"That":[161],"collapse":[162],"warns":[163],"this":[164],"line":[165],"work":[167],"downstream":[169],"score":[171,187],"cannot":[172],"validate":[173],"self-evolved":[175],"evaluator,":[176],"since":[177],"collapsed":[179],"trains":[181],"skills":[182,226],"just":[183],"as":[184],"well.":[185],"Task":[186],"answers":[188],"only":[189],"sufficiency,":[190],"evolved":[193,225],"suffices:":[195],"\\emph{Double":[196],"Ratchet},":[197],"co-evolving":[198],"lifecycle-managed":[203],"loop,":[205],"retains":[206],"88--110\\%":[207],"lift":[210],"or":[213],"hand-written":[215],"rubric":[216],"buys,":[217],"across":[218],"MBPP+,":[219],"Spider~2.0-Snow,":[220],"report":[222,229],"generation.":[223],"When":[224],"gamed":[227],"rubric,":[230],"independent":[232],"caught":[234],"added":[238],"repaired":[240],"it.":[241]},"counts_by_year":[],"updated_date":"2026-08-01T09:00:35.917206","created_date":"2026-07-16T00:00:00"}