{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BGRULHLZNUUTXKJZINGIKWOBDJ","short_pith_number":"pith:BGRULHLZ","schema_version":"1.0","canonical_sha256":"09a3459d796d293ba939434c8559c11a46b2ef430c2cee29ee7b4ec5dfacad02","source":{"kind":"arxiv","id":"2605.19999","version":1},"attestation_state":"computed","paper":{"title":"LLM Benchmark Datasets Should Be Contamination-Resistant","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.LG","authors_text":"Ali Al-Lawati, Dongwon Lee, Jason Lucas, Suhang Wang","submitted_at":"2026-05-19T15:33:16Z","abstract_excerpt":"Benchmark datasets are critical for reproducible, reliable, and discriminative evaluation of LLMs. However, recent studies reveal that many benchmark datasets are included in pretraining corpora, i.e., $\\textit{contaminated}$, which diminishes their value as reliable measures of model generalization. In this paper, we argue that benchmark datasets should be $\\textit{contamination-resistant}$, i.e., $\\textit{unlearnable}$, but support $\\textit{inference}$. To accomplish this, we first highlight the wide prevalence of benchmark dataset contamination and outline the properties of contamination-re"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.19999","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-19T15:33:16Z","cross_cats_sorted":["cs.AI","cs.CR"],"title_canon_sha256":"059ead94e76a7e85633f5d4b145c050d2fd31b851101920584dd39d41e227ca1","abstract_canon_sha256":"6dda7a0559621e1148817af99ac77cf224857ec31db23d34d62502a40cf6dd1e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:05:58.551443Z","signature_b64":"7vD7GbFkjU5Mp0EwotcMOOc+IIFQ9hPwUQLq/GZHRkIvAh52K6VZQeatQ3poIiT9iwPSohOsJawkdT5IgDeqCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"09a3459d796d293ba939434c8559c11a46b2ef430c2cee29ee7b4ec5dfacad02","last_reissued_at":"2026-05-20T02:05:58.550948Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:05:58.550948Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLM Benchmark Datasets Should Be Contamination-Resistant","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.LG","authors_text":"Ali Al-Lawati, Dongwon Lee, Jason Lucas, Suhang Wang","submitted_at":"2026-05-19T15:33:16Z","abstract_excerpt":"Benchmark datasets are critical for reproducible, reliable, and discriminative evaluation of LLMs. However, recent studies reveal that many benchmark datasets are included in pretraining corpora, i.e., $\\textit{contaminated}$, which diminishes their value as reliable measures of model generalization. In this paper, we argue that benchmark datasets should be $\\textit{contamination-resistant}$, i.e., $\\textit{unlearnable}$, but support $\\textit{inference}$. To accomplish this, we first highlight the wide prevalence of benchmark dataset contamination and outline the properties of contamination-re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19999","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.19999/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.19999","created_at":"2026-05-20T02:05:58.551019+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19999v1","created_at":"2026-05-20T02:05:58.551019+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19999","created_at":"2026-05-20T02:05:58.551019+00:00"},{"alias_kind":"pith_short_12","alias_value":"BGRULHLZNUUT","created_at":"2026-05-20T02:05:58.551019+00:00"},{"alias_kind":"pith_short_16","alias_value":"BGRULHLZNUUTXKJZ","created_at":"2026-05-20T02:05:58.551019+00:00"},{"alias_kind":"pith_short_8","alias_value":"BGRULHLZ","created_at":"2026-05-20T02:05:58.551019+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ","json":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ.json","graph_json":"https://pith.science/api/pith-number/BGRULHLZNUUTXKJZINGIKWOBDJ/graph.json","events_json":"https://pith.science/api/pith-number/BGRULHLZNUUTXKJZINGIKWOBDJ/events.json","paper":"https://pith.science/paper/BGRULHLZ"},"agent_actions":{"view_html":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ","download_json":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ.json","view_paper":"https://pith.science/paper/BGRULHLZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19999&json=true","fetch_graph":"https://pith.science/api/pith-number/BGRULHLZNUUTXKJZINGIKWOBDJ/graph.json","fetch_events":"https://pith.science/api/pith-number/BGRULHLZNUUTXKJZINGIKWOBDJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ/action/storage_attestation","attest_author":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ/action/author_attestation","sign_citation":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ/action/citation_signature","submit_replication":"https://pith.science/pith/BGRULHLZNUUTXKJZINGIKWOBDJ/action/replication_record"}},"created_at":"2026-05-20T02:05:58.551019+00:00","updated_at":"2026-05-20T02:05:58.551019+00:00"}