{"id":"https://openalex.org/W7159581242","doi":"https://doi.org/10.48550/arxiv.2604.27547","title":"Diagnosing Capability Gaps in Fine-Tuning Data","display_name":"Diagnosing Capability Gaps in Fine-Tuning Data","publication_year":2026,"publication_date":"2026-04-30","ids":{"openalex":"https://openalex.org/W7159581242","doi":"https://doi.org/10.48550/arxiv.2604.27547"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.27547","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.27547","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.2604.27547","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5056793060","display_name":"Saied Asgari Taghanaki","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Taghanaki, Saeid Asgari","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048964944","display_name":"Rakshanda Agarwal","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Agarwal, Rakshanda","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047821791","display_name":"Bruce Sun","orcid":"https://orcid.org/0000-0002-1714-0208"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Bruce","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134976655","display_name":"Rohan Jha","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jha, Rohan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5038276570","display_name":"Elias Stengel-Eskin","orcid":"https://orcid.org/0000-0002-6689-505X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Stengel-Eskin, Elias","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006484950","display_name":"Sara Malvar","orcid":"https://orcid.org/0000-0002-3209-4588"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Malvar, Sara","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134933742","display_name":"Rui Ying","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ying, Rui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134961530","display_name":"Yifei Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yifei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074731322","display_name":"Guilherme Potje","orcid":"https://orcid.org/0000-0003-2577-2886"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Potje, Guilherme","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043473455","display_name":"Tusher Chakraborty","orcid":"https://orcid.org/0000-0003-1656-5471"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chakraborty, Tusher","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126845177","display_name":"Leonardo de Oliveira Nunes","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nunes, Leonardo de Oliveira","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112443217","display_name":"Ranveer Chandra","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chandra, Ranveer","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134985993","display_name":"Emre Kiciman","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kiciman, Emre","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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.5213000178337097,"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"}},"topics":[{"id":"https://openalex.org/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.5213000178337097,"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"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.09040000289678574,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.08780000358819962,"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/sample","display_name":"Sample (material)","score":0.661899983882904},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.6243000030517578},{"id":"https://openalex.org/keywords/decomposition","display_name":"Decomposition","score":0.5335000157356262},{"id":"https://openalex.org/keywords/closing","display_name":"Closing (real estate)","score":0.5044999718666077},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.4731999933719635},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.4052000045776367},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.38199999928474426}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7620999813079834},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.661899983882904},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6243000030517578},{"id":"https://openalex.org/C124681953","wikidata":"https://www.wikidata.org/wiki/Q339062","display_name":"Decomposition","level":2,"score":0.5335000157356262},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.506600022315979},{"id":"https://openalex.org/C2778775528","wikidata":"https://www.wikidata.org/wiki/Q5135432","display_name":"Closing (real estate)","level":2,"score":0.5044999718666077},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.48500001430511475},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.4821000099182129},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.4731999933719635},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.4052000045776367},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.38199999928474426},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.3806000053882599},{"id":"https://openalex.org/C2780428219","wikidata":"https://www.wikidata.org/wiki/Q16952335","display_name":"Cover (algebra)","level":2,"score":0.35989999771118164},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.3188999891281128},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.296099990606308},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.29429998993873596},{"id":"https://openalex.org/C9357733","wikidata":"https://www.wikidata.org/wiki/Q6878417","display_name":"Missing data","level":2,"score":0.29330000281333923},{"id":"https://openalex.org/C3020318244","wikidata":"https://www.wikidata.org/wiki/Q4812187","display_name":"Large sample","level":2,"score":0.2630999982357025}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.27547","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.27547","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.2604.27547","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.27547","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":0.7915135622024536,"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Fine-tuning":[0],"large":[1],"language":[2],"models":[3],"(LLMs)":[4],"for":[5,145,217],"domain-specific":[6],"tasks":[7],"requires":[8],"training":[9,83,164],"datasets":[10,56],"that":[11,47,127,201],"comprehensively":[12],"cover":[13],"the":[14,105,169,180,192],"target":[15,136],"capabilities":[16,23,95],"a":[17,24,38,45,67,73,156,205],"practitioner":[18,68],"needs.":[19],"Yet":[20],"identifying":[21],"which":[22],"dataset":[25],"fails":[26],"to":[27,174],"support,":[28],"and":[29,61,92,183,213],"doing":[30],"so":[31],"before":[32],"an":[33,85],"expensive":[34],"fine-tuning":[35,55],"run,":[36],"remains":[37],"largely":[39],"unsolved":[40],"problem.":[41],"We":[42,103],"introduce":[43],"GoalCover,":[44],"framework":[46,106],"helps":[48],"practitioners":[49],"systematically":[50],"detect":[51],"capability":[52,134,211],"gaps":[53,212],"in":[54],"through":[57,69,96,112],"interactive":[58],"goal":[59,75],"decomposition":[60,71],"automated":[62,97],"coverage":[63],"assessment.":[64],"GoalCover":[65,128,202],"guides":[66],"structured":[70],"of":[72,99,177],"high-level":[74],"into":[76],"atomic,":[77],"independently":[78],"evaluable":[79],"subgoals;":[80],"assigns":[81],"each":[82],"sample":[84,101],"LLM-based":[86],"alignment":[87],"score":[88],"against":[89],"every":[90],"subgoal;":[91],"surfaces":[93],"missing":[94],"analysis":[98],"low-scoring":[100],"explanations.":[102],"validate":[104],"along":[107],"two":[108,197],"complementary":[109],"axes.":[110],"First,":[111],"controlled":[113],"corruption":[114],"experiments":[115],"across":[116],"three":[117],"domains":[118],"(medical":[119],"QA,":[120],"legal":[121],"summarization,":[122],"code":[123],"generation),":[124],"we":[125,151],"show":[126,200],"reliably":[129],"distinguishes":[130],"targeted":[131],"from":[132,172],"non-targeted":[133],"impacts:":[135],"subgoals":[137,147],"degrade":[138],"by":[139],"25.6%":[140],"on":[141,155,165],"average":[142],"versus":[143],"2.1%":[144],"non-target":[146],"(Cohen's":[148],"d=1.24).":[149],"Second,":[150],"demonstrate":[152],"downstream":[153],"utility":[154],"financial-summarization":[157],"Reinforcement":[158],"Fine-Tuning":[159],"(RFT)":[160],"task":[161],"with":[162,187],"Qwen-3-14B:":[163],"GoalCover-filtered":[166],"data":[167,186],"improves":[168],"LLM-judge":[170],"reward":[171],"3.77":[173],"4.12":[175],"(out":[176],"5)":[178],"over":[179],"unfiltered":[181],"baseline,":[182],"combining":[184],"filtered":[185],"goal-conditioned":[188],"synthetic":[189],"samples":[190],"yields":[191],"strongest":[193],"result":[194],"(4.20).":[195],"The":[196],"results":[198],"together":[199],"works":[203],"as":[204],"practical":[206],"pre-fine-tuning":[207],"diagnostic:":[208],"it":[209],"detects":[210],"produces":[214],"concrete":[215],"signal":[216],"closing":[218],"them.":[219]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-02T00:00:00"}