{"id":"https://openalex.org/W7148457573","doi":"https://doi.org/10.48550/arxiv.2604.01025","title":"Fast and Accurate Probing of In-Training LLMs' Downstream Performances","display_name":"Fast and Accurate Probing of In-Training LLMs' Downstream Performances","publication_year":2026,"publication_date":"2026-04-01","ids":{"openalex":"https://openalex.org/W7148457573","doi":"https://doi.org/10.48550/arxiv.2604.01025"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.01025","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.01025","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.2604.01025","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Liu, Zhichen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Zhichen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132818158","display_name":"Tianle Lun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lun, Tianle","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132755301","display_name":"Zhibin Wen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wen, Zhibin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112372021","display_name":"Hao An","orcid":"https://orcid.org/0009-0003-5256-0878"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"An, Hao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132750485","display_name":"Yulin Ou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ou, Yulin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037525206","display_name":"J. Xu","orcid":"https://orcid.org/0000-0001-7597-2179"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Jianhui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132797780","display_name":"Hao Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Hao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029831702","display_name":"Wenyi Fang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fang, Wenyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132824522","display_name":"Yang Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5132809818","display_name":"Yang Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yang","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/T10028","display_name":"Topic Modeling","score":0.3206999897956848,"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":0.3206999897956848,"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":0.24799999594688416,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.0617000013589859,"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/downstream","display_name":"Downstream (manufacturing)","score":0.8628000020980835},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.5909000039100647},{"id":"https://openalex.org/keywords/generalizability-theory","display_name":"Generalizability theory","score":0.5561000108718872},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5249999761581421},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5110999941825867},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.4652999937534332},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.4507000148296356},{"id":"https://openalex.org/keywords/generative-grammar","display_name":"Generative grammar","score":0.3790000081062317}],"concepts":[{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.8628000020980835},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7364000082015991},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5909000039100647},{"id":"https://openalex.org/C27158222","wikidata":"https://www.wikidata.org/wiki/Q5532422","display_name":"Generalizability theory","level":2,"score":0.5561000108718872},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5249999761581421},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5110999941825867},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.4652999937534332},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.4507000148296356},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.3790000081062317},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.33719998598098755},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.33559998869895935},{"id":"https://openalex.org/C169903167","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Test set","level":2,"score":0.3328999876976013},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.3312000036239624},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.32580000162124634},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.31779998540878296},{"id":"https://openalex.org/C200601418","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability engineering","level":1,"score":0.3174000084400177},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.29600000381469727},{"id":"https://openalex.org/C2777115002","wikidata":"https://www.wikidata.org/wiki/Q7168246","display_name":"Performance prediction","level":2,"score":0.28690001368522644},{"id":"https://openalex.org/C191172861","wikidata":"https://www.wikidata.org/wiki/Q7899321","display_name":"Upstream (networking)","level":2,"score":0.2815999984741211},{"id":"https://openalex.org/C2778496695","wikidata":"https://www.wikidata.org/wiki/Q254128","display_name":"Dilemma","level":2,"score":0.27549999952316284},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.2718000113964081},{"id":"https://openalex.org/C128942645","wikidata":"https://www.wikidata.org/wiki/Q1568346","display_name":"Test case","level":3,"score":0.25859999656677246},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.2533999979496002}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.01025","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.01025","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.2604.01025","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.01025","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0,107,154],"paradigm":[1,32,97],"of":[2,19,26,39,113,151],"scaling":[3],"Large":[4],"Language":[5],"Models":[6],"(LLMs)":[7],"in":[8,83],"both":[9],"parameter":[10],"size":[11],"and":[12,80,120,173,196,207],"test":[13],"time":[14],"has":[15],"pushed":[16],"the":[17,24,28,37,66,110,123,144,175],"boundaries":[18],"AI":[20],"capabilities,":[21],"but":[22],"at":[23],"cost":[25],"making":[27,36],"traditional":[29],"generative":[30,183],"evaluation":[31,44,96,184,200],"prohibitively":[33],"expensive,":[34],"therefore":[35],"latency":[38,177],"LLM's":[40],"in-training":[41,95,198],"downstream":[42,58,105,127,152,199],"performance":[43,125,161],"unbearable.":[45],"However,":[46],"simple":[47],"metrics":[48],"like":[49],"training":[50],"loss":[51],"(perplexity)":[52],"are":[53],"not":[54],"always":[55],"correlated":[56],"with":[57],"performance,":[59],"as":[60,118],"sometimes":[61],"their":[62,141],"trends":[63],"diverge":[64],"from":[65,178],"actual":[67],"task":[68],"outcomes.":[69],"This":[70],"dilemma":[71],"calls":[72],"for":[73,103],"a":[74,93,100,148,159,194,203],"method":[75],"that":[76,98],"is":[77],"computationally":[78],"efficient":[79,208],"sufficiently":[81],"accurate":[82],"measuring":[84],"model":[85],"capabilities.":[86],"To":[87],"address":[88],"this":[89,191],"challenge,":[90],"we":[91],"introduce":[92],"new":[94],"uses":[99],"lightweight":[101],"probe":[102,138],"monitoring":[104],"performance.":[106],"probes":[108,155],"take":[109],"internal":[111],"representations":[112],"LLM":[114,209],"checkpoints":[115,146,169],"(during":[116],"training)":[117],"input":[119],"directly":[121],"predict":[122,158],"checkpoint's":[124,160],"on":[126],"tasks":[128],"measured":[129],"by":[130],"success":[131],"probability":[132],"(i.e.,":[133],"pass@1).":[134],"We":[135],"design":[136],"several":[137],"architectures,":[139],"validating":[140],"effectiveness":[142],"using":[143],"OLMo3-7B's":[145],"across":[147,168],"diverse":[149],"set":[150],"tasks.":[153],"can":[156],"accurately":[157],"(with":[162],"avg.":[163],"AUROC$>$0.75),":[164],"have":[165],"decent":[166],"generalizability":[167],"(earlier":[170],"predicts":[171],"later),":[172],"reduce":[174],"computation":[176],"$\\sim$1":[179],"hr":[180],"(using":[181],"conventional":[182],"method)":[185],"to":[186],"$\\sim$3":[187],"min.":[188],"In":[189],"sum,":[190],"work":[192],"presents":[193],"practical":[195],"scalable":[197],"paradigm,":[201],"enabling":[202],"more":[204],"agile,":[205],"informed,":[206],"development":[210],"process.":[211]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-03T00:00:00"}