{"id":"https://openalex.org/W7161769046","doi":"https://doi.org/10.48550/arxiv.2605.19999","title":"LLM Benchmark Datasets Should Be Contamination-Resistant","display_name":"LLM Benchmark Datasets Should Be Contamination-Resistant","publication_year":2026,"publication_date":"2026-05-19","ids":{"openalex":"https://openalex.org/W7161769046","doi":"https://doi.org/10.48550/arxiv.2605.19999"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.19999","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19999","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":"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.2605.19999","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5127889923","display_name":"Ali Al-Lawati","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Al-Lawati, Ali","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136532039","display_name":"Jason Lucas","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lucas, Jason","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136559685","display_name":"Dongwon Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Dongwon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136590001","display_name":"Suhang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Suhang","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":0.2558000087738037,"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":0.2558000087738037,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.1266999989748001,"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":0.09380000084638596,"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/benchmarking","display_name":"Benchmarking","score":0.8446000218391418},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.8263999819755554},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6290000081062317},{"id":"https://openalex.org/keywords/discriminative-model","display_name":"Discriminative model","score":0.5929999947547913},{"id":"https://openalex.org/keywords/interoperability","display_name":"Interoperability","score":0.4562999904155731},{"id":"https://openalex.org/keywords/reliability","display_name":"Reliability (semiconductor)","score":0.36730000376701355}],"concepts":[{"id":"https://openalex.org/C86251818","wikidata":"https://www.wikidata.org/wiki/Q816754","display_name":"Benchmarking","level":2,"score":0.8446000218391418},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.8263999819755554},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7648000121116638},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6290000081062317},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.5929999947547913},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5687999725341797},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.5371000170707703},{"id":"https://openalex.org/C20136886","wikidata":"https://www.wikidata.org/wiki/Q749647","display_name":"Interoperability","level":2,"score":0.4562999904155731},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4404999911785126},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.36730000376701355},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.2863999903202057},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.2750999927520752},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.2711000144481659},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2605000138282776},{"id":"https://openalex.org/C63000827","wikidata":"https://www.wikidata.org/wiki/Q3080428","display_name":"Software portability","level":2,"score":0.25519999861717224}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.19999","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19999","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":"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.2605.19999","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19999","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":"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":[{"display_name":"Reduced inequalities","id":"https://metadata.un.org/sdg/10","score":0.6319249868392944}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Benchmark":[0],"datasets":[1,19,44,103],"are":[2,20],"critical":[3],"for":[4],"reproducible,":[5],"reliable,":[6],"and":[7,66,82,135,137],"discriminative":[8],"evaluation":[9,144],"of":[10,34,62,70,122],"LLMs.":[11],"However,":[12],"recent":[13],"studies":[14],"reveal":[15],"that":[16,42],"many":[17],"benchmark":[18,43,63],"included":[21],"in":[22,85],"pretraining":[23],"corpora,":[24],"i.e.,":[25,48],"$\\textit{contaminated}$,":[26],"which":[27],"diminishes":[28],"their":[29],"value":[30],"as":[31],"reliable":[32],"measures":[33],"model":[35],"generalization.":[36],"In":[37],"this":[38],"paper,":[39],"we":[40,56,74,96,113],"argue":[41],"should":[45],"be":[46,90],"$\\textit{contamination-resistant}$,":[47],"$\\textit{unlearnable}$,":[49],"but":[50],"support":[51,93],"$\\textit{inference}$.":[52],"To":[53],"accomplish":[54],"this,":[55],"first":[57],"highlight":[58,75],"the":[59,68,77,80,86,111,116,120],"wide":[60],"prevalence":[61],"dataset":[64],"contamination":[65],"outline":[67,97],"properties":[69],"contamination-resistant":[71,129,140],"datasets.":[72],"Second,":[73],"how":[76],"asymmetry":[78],"between":[79],"inference":[81],"training":[83],"pipelines":[84],"Transformer":[87],"architecture":[88],"can":[89],"leveraged":[91],"to":[92,100,118],"contamination-resistance.":[94],"Third,":[95],"mathematical":[98],"advancements":[99],"make":[101],"these":[102],"interoperable":[104],"across":[105],"various":[106],"LLM":[107,123],"architectures.":[108],"Based":[109],"on":[110,115],"above,":[112],"call":[114],"community":[117],"ensure":[119],"reliability":[121],"benchmarking":[124],"by:":[125],"(i)":[126],"advancing":[127],"novel":[128],"methodologies,":[130],"(ii)":[131],"developing":[132],"supporting":[133],"methods":[134],"platforms,":[136],"(iii)":[138],"adopting":[139],"benchmarks":[141],"into":[142],"existing":[143],"pipelines.":[145]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-21T00:00:00"}