{"id":"https://openalex.org/W7165636148","doi":"https://doi.org/10.48550/arxiv.2606.20657","title":"A-Evolve-Training: Autonomous Post-Training of a 30B Model","display_name":"A-Evolve-Training: Autonomous Post-Training of a 30B Model","publication_year":2026,"publication_date":"2026-06-09","ids":{"openalex":"https://openalex.org/W7165636148","doi":"https://doi.org/10.48550/arxiv.2606.20657"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.20657","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.20657","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":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.20657","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139183434","display_name":"Zhan Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Zhan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139216955","display_name":"Bing He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Bing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078860840","display_name":"Yisi Sang","orcid":"https://orcid.org/0000-0002-8876-7542"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sang, Yisi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029653356","display_name":"H\u00fcseyin Ozan \u04aairkino\u011flu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu, Hanqing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":null,"display_name":"Dumoulin, Benoit","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dumoulin, Benoit","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":0.09040000289678574,"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":0.09040000289678574,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.059700001031160355,"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":0.054099999368190765,"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/metric","display_name":"Metric (unit)","score":0.5540000200271606},{"id":"https://openalex.org/keywords/autonomous-system","display_name":"Autonomous system (mathematics)","score":0.5210000276565552},{"id":"https://openalex.org/keywords/human-in-the-loop","display_name":"Human-in-the-loop","score":0.460999995470047},{"id":"https://openalex.org/keywords/frame","display_name":"Frame (networking)","score":0.4440999925136566},{"id":"https://openalex.org/keywords/domain","display_name":"Domain (mathematical analysis)","score":0.439300000667572},{"id":"https://openalex.org/keywords/proxy","display_name":"Proxy (statistics)","score":0.4392000138759613},{"id":"https://openalex.org/keywords/reading","display_name":"Reading (process)","score":0.4059000015258789},{"id":"https://openalex.org/keywords/tracking-system","display_name":"Tracking system","score":0.37950000166893005}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6646000146865845},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.5540000200271606},{"id":"https://openalex.org/C9628104","wikidata":"https://www.wikidata.org/wiki/Q788009","display_name":"Autonomous system (mathematics)","level":2,"score":0.5210000276565552},{"id":"https://openalex.org/C2780626000","wikidata":"https://www.wikidata.org/wiki/Q5936775","display_name":"Human-in-the-loop","level":2,"score":0.460999995470047},{"id":"https://openalex.org/C126042441","wikidata":"https://www.wikidata.org/wiki/Q1324888","display_name":"Frame (networking)","level":2,"score":0.4440999925136566},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.439300000667572},{"id":"https://openalex.org/C2780148112","wikidata":"https://www.wikidata.org/wiki/Q1432581","display_name":"Proxy (statistics)","level":2,"score":0.4392000138759613},{"id":"https://openalex.org/C554936623","wikidata":"https://www.wikidata.org/wiki/Q199657","display_name":"Reading (process)","level":2,"score":0.4059000015258789},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3952000141143799},{"id":"https://openalex.org/C154586513","wikidata":"https://www.wikidata.org/wiki/Q4420972","display_name":"Tracking system","level":3,"score":0.37950000166893005},{"id":"https://openalex.org/C2775936607","wikidata":"https://www.wikidata.org/wiki/Q466845","display_name":"Tracking (education)","level":2,"score":0.36079999804496765},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.3465999960899353},{"id":"https://openalex.org/C512654426","wikidata":"https://www.wikidata.org/wiki/Q19652","display_name":"Public domain","level":2,"score":0.32499998807907104},{"id":"https://openalex.org/C158154518","wikidata":"https://www.wikidata.org/wiki/Q7310970","display_name":"Relevance (law)","level":2,"score":0.2953000068664551},{"id":"https://openalex.org/C207685749","wikidata":"https://www.wikidata.org/wiki/Q2088941","display_name":"Domain knowledge","level":2,"score":0.27970001101493835},{"id":"https://openalex.org/C13687954","wikidata":"https://www.wikidata.org/wiki/Q4826847","display_name":"Autonomous agent","level":2,"score":0.27959999442100525},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.27160000801086426},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.2676999866962433},{"id":"https://openalex.org/C115901376","wikidata":"https://www.wikidata.org/wiki/Q184199","display_name":"Automation","level":2,"score":0.2612000107765198},{"id":"https://openalex.org/C184670325","wikidata":"https://www.wikidata.org/wiki/Q512604","display_name":"Loop (graph theory)","level":2,"score":0.25049999356269836}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.20657","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.20657","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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.20657","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.20657","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":"Preprint"},"sustainable_development_goals":[{"score":0.5654082298278809,"id":"https://metadata.un.org/sdg/9","display_name":"Industry, innovation and infrastructure"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Post-training":[0],"a":[1,39,53,146,190,206,287],"frontier":[2],"model":[3,51],"is":[4,194,217,225,276],"normally":[5],"weeks":[6],"of":[7,56,73,78,189,197,210],"human":[8,34,61,211,259,289],"work:":[9],"proposing":[10],"data":[11],"and":[12,115,165,219,253],"recipe":[13],"changes,":[14],"launching":[15],"runs,":[16],"reading":[17],"evals,":[18],"deciding":[19],"what":[20,167],"to":[21,106,221],"keep.":[22],"We":[23,138,171,202],"report":[24],"an":[25],"autonomous":[26,148,208,230],"system":[27,178,248],"that":[28,88,128,145,158,176,198,265,270,273],"runs":[29],"this":[30,140,193,224,234,262],"loop":[31,86,149,267],"with":[32,256,281],"no":[33,121,257],"in":[35],"the":[36,59,65,76,83,85,99,111,130,135,173,180,226,251,266,282],"loop,":[37],"post-training":[38,188,231],"30B":[40],"Nemotron":[41],"across":[42],"four":[43],"rounds":[44],"over":[45],"multiple":[46],"weeks.":[47],"The":[48,213,246],"autonomously":[49],"produced":[50],"reaches":[52],"held-out":[54],"score":[55],"0.86":[57],"against":[58],"top":[60],"submission's":[62],"0.87":[63],"on":[64,98],"public":[66,238,258],"NVIDIA":[67],"Nemotron-Reasoning":[68],"Challenge":[69],"leaderboard,":[70],"placing":[71],"8th":[72],"~4000":[74],"at":[75,233,242,269],"time":[77],"writing.":[79],"More":[80],"striking":[81],"than":[82],"number:":[84],"detected":[87,157],"its":[89,117,159,274],"own":[90,118],"dev":[91,105,124],"metric":[92],"had":[93,162],"stopped":[94],"tracking":[95],"external":[96,112,136],"performance":[97],"weakest":[100],"domain":[101],"--":[102,114,278],"candidates":[103],"drove":[104],"record":[107],"highs":[108],"without":[109],"moving":[110],"target":[113],"revised":[116],"search":[119],"policy,":[120],"longer":[122],"maximizing":[123],"but":[125],"seeking":[126],"interventions":[127],"lowered":[129],"now-misleading":[131],"proxy":[132],"while":[133],"improving":[134],"target.":[137],"treat":[139],"as":[141,169],"direct,":[142],"auditable":[143],"evidence":[144],"scaled":[147],"can":[150],"produce":[151],"discovery,":[152],"not":[153,204,272],"only":[154,264],"optimization:":[155],"it":[156],"measurement":[160],"frame":[161],"become":[163],"misleading":[164],"changed":[166],"counted":[168],"evidence.":[170],"take":[172],"operational":[174],"view":[175],"any":[177],"worth":[179],"\"recursive":[181],"self-improvement\"":[182],"label":[183],"must":[184],"eventually":[185],"perform":[186],"end-to-end":[187],"frontier-class":[191],"model;":[192],"one":[195],"datapoint":[196],"bar":[199],"being":[200],"cleared.":[201],"do":[203],"claim":[205,214,284],"\"first":[207],"match\"":[209],"researchers.":[212],"we":[215],"make":[216],"narrower":[218],"auditable:":[220],"our":[222],"knowledge,":[223],"first":[227],"publicly":[228],"reported":[229],"run":[232],"scale,":[235,271],"where":[236],"prior":[237],"autonomous-ML-research":[239],"demonstrations":[240],"sit":[241],"GPT-2-class":[243],"(~124M)":[244],"budgets.":[245],"same":[247],"also":[249],"post-trains":[250],"120B":[252],"550B":[254],"Nemotron;":[255],"baseline":[260],"there,":[261],"shows":[263],"closes":[268],"output":[275],"competitive":[277],"infrastructure":[279],"evidence,":[280],"effectiveness":[283],"deferred":[285],"until":[286],"comparable":[288],"anchor":[290],"exists.":[291]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-24T00:00:00"}