{"id":"https://openalex.org/W7213675664","doi":null,"title":"Form Over Content In Gradient-Based Data Attribution Methods","display_name":"Form Over Content In Gradient-Based Data Attribution Methods","publication_year":2026,"publication_date":"2026-09-17","ids":{"openalex":"https://openalex.org/W7213675664"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.19589","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.19589","pdf_url":"https://arxiv.org/pdf/2609.19589","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.19589","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5152154354","display_name":"Sunwoo Kim","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Sunwoo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136072028","display_name":"Seokwon Jung","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jung, Seokwon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5152154206","display_name":"Sohyung Kim","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Sohyung","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120057418","display_name":"Seong Joon Oh","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Oh, Seong Joon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5122350367","display_name":"Alice Oh","orcid":"https://orcid.org/0000-0002-7884-3038","observed_orcids":["https://orcid.org/0000-0002-7884-3038"]},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Oh, Alice","raw_affiliation_strings":["KAIST"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"KAIST","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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":0.5371000170707703,"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":0.5371000170707703,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.2786000072956085,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.050200000405311584,"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/gradient-similarity","display_name":"gradient similarity","score":0.902999997138977},{"id":"https://openalex.org/keywords/answer-format","display_name":"answer format","score":0.8360000252723694},{"id":"https://openalex.org/keywords/dataset-selection","display_name":"data selection","score":0.5699999928474426},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.7400000095367432},{"id":"https://openalex.org/keywords/supervised-fine-tuning","display_name":"supervised fine-tuning","score":0.8339999914169312},{"id":"https://openalex.org/keywords/less","display_name":"LESS","score":0.43299999833106995},{"id":"https://openalex.org/keywords/instruction-tuning","display_name":"instruction tuning","score":0.6940000057220459},{"id":"https://openalex.org/keywords/benchmark-dataset","display_name":"benchmark dataset","score":0.36399999260902405},{"id":"https://openalex.org/keywords/interpretability","display_name":"interpretability","score":0.414000004529953}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6830000281333923},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.6241000294685364},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5952000021934509},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.588699996471405},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5580999851226807},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.5116000175476074},{"id":"https://openalex.org/C130318100","wikidata":"https://www.wikidata.org/wiki/Q2268914","display_name":"Semantic similarity","level":2,"score":0.5091000199317932},{"id":"https://openalex.org/C2780762811","wikidata":"https://www.wikidata.org/wiki/Q1784941","display_name":"Cosine similarity","level":3,"score":0.4700999855995178},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.45809999108314514},{"id":"https://openalex.org/C143299363","wikidata":"https://www.wikidata.org/wiki/Q900584","display_name":"Attribution","level":2,"score":0.4399999976158142},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.41929998993873596},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.367900013923645},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.3675000071525574},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3549000024795532},{"id":"https://openalex.org/C2780876879","wikidata":"https://www.wikidata.org/wiki/Q3054749","display_name":"Meaning (existential)","level":2,"score":0.3163999915122986},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.30880001187324524},{"id":"https://openalex.org/C3020202489","wikidata":"https://www.wikidata.org/wiki/Q2032038","display_name":"Authorship attribution","level":2,"score":0.3009999990463257},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.3005000054836273},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.25760000944137573},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.2567000091075897}],"mesh":[],"locations_count":1,"locations":[{"id":"pmh:oai:arXiv.org:2609.19589","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.19589","pdf_url":"https://arxiv.org/pdf/2609.19589","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.19589","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.19589","pdf_url":"https://arxiv.org/pdf/2609.19589","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7213675664.pdf","grobid_xml":"https://content.openalex.org/works/W7213675664.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4411120136","https://openalex.org/W7220465494","https://openalex.org/W4405094466","https://openalex.org/W7163720809","https://openalex.org/W7215041697","https://openalex.org/W4414693908","https://openalex.org/W7125102852","https://openalex.org/W4412945156","https://openalex.org/W7154941062","https://openalex.org/W4404261169"],"abstract_inverted_index":{"Data":[0],"attribution":[1,172],"methods":[2,173],"using":[3],"gradient":[4,21,87],"similarity":[5,22,176],"are":[6],"widely":[7],"used":[8],"to":[9],"analyze":[10,141],"and":[11,56,133,137,155,202,209],"select":[12],"training":[13],"data":[14,149,198],"for":[15,49,152,206],"large":[16],"language":[17],"models,":[18],"but":[19],"what":[20],"actually":[23],"measures":[24],"is":[25,41],"debated.":[26],"Some":[27],"interpret":[28],"it":[29],"as":[30,93,185,187],"identifying":[31],"task-relevant":[32],"skills,":[33],"while":[34,106],"other":[35],"work":[36],"reports":[37],"that":[38,69,86,122,157,170,182],"surface":[39],"form":[40],"the":[42,90,127,142,162,188,192],"main":[43],"factor.":[44],"We":[45,84,120,139],"resolve":[46],"this":[47,123],"debate":[48],"supervised":[50],"fine-tuning":[51],"examples":[52],"by":[53],"varying":[54],"task":[55,74,179,203],"answer":[57,66,91,98,112,165,200],"format":[58,77,80,99,113,175,201],"independently.":[59],"Specifically,":[60],"we":[61,168],"render":[62],"benchmarks":[63,108],"in":[64],"different":[65,111],"formats,":[67],"such":[68,183],"datasets":[70],"can":[71],"share":[72],"a":[73,76,79,82,147],"without":[75,81],"or":[78],"task.":[83],"find":[85,156],"alignment":[88,117],"follows":[89],"format,":[92],"benchmark":[94],"pairs":[95],"sharing":[96],"an":[97],"align":[100],"strongly":[101],"(disattenuated":[102],"cosine":[103],"near":[104],"0.4),":[105],"same":[107],"rendered":[109],"with":[110],"classes":[114],"show":[115],"no":[116],"(near":[118],"0.0).":[119],"demonstrate":[121,169],"ordering":[124],"holds":[125],"from":[126],"earliest":[128],"pretraining":[129],"checkpoints":[130],"through":[131],"post-training,":[132],"across":[134],"model":[135],"scales":[136],"families.":[138],"then":[140],"released":[143],"selections":[144,160],"of":[145,191],"LESS,":[146],"gradient-based":[148,171],"selection":[150],"method":[151],"instruction":[153],"tuning,":[154],"each":[158],"target's":[159,163],"over-represent":[161],"own":[164],"format.":[166],"Hence,":[167],"track":[174],"more":[177],"than":[178],"semantics,":[180],"meaning":[181],"methods,":[184],"well":[186],"semantic":[189],"interpretation":[190],"gradient,":[193],"should":[194],"be":[195],"tested":[196],"on":[197],"where":[199],"vary":[204],"independently":[205],"greater":[207],"robustness":[208],"reliability.":[210]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-20T00:00:00"}