{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IHBV4OAHLDTPDYXK3QBXIBXJ6K","short_pith_number":"pith:IHBV4OAH","schema_version":"1.0","canonical_sha256":"41c35e380758e6f1e2eadc037406e9f2933e80625681aba0dcfe25638fd3813c","source":{"kind":"arxiv","id":"2402.02823","version":2},"attestation_state":"computed","paper":{"title":"Evading Data Contamination Detection for Language Models is (too) Easy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Jasper Dekoninck, Marc Fischer, Mark Niklas M\\\"uller, Martin Vechev, Maximilian Baader","submitted_at":"2024-02-05T09:10:32Z","abstract_excerpt":"Large language models are widespread, with their performance on benchmarks frequently guiding user preferences for one model over another. However, the vast amount of data these models are trained on can inadvertently lead to contamination with public benchmarks, thus compromising performance measurements. While recently developed contamination detection methods try to address this issue, they overlook the possibility of deliberate contamination by malicious model providers aiming to evade detection. We argue that this setting is of crucial importance as it casts doubt on the reliability of pu"},"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":"2402.02823","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-05T09:10:32Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CR"],"title_canon_sha256":"12b676ecb158be7ac53c672243d8cca0d0b34bcf2f2130373f5f6699a264cbd8","abstract_canon_sha256":"e5b55e679610037e8ba309ed7c3784e86b667d14bab4a479b56f8aad06e3176d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:44:00.358010Z","signature_b64":"e+VJQdTbrcA4XEsyymswwK0zJAoZRtwzwHCBNnN7DuYFKZommCvPtYfRagS71pv7uRftPsTD0bXIF/SbltiQBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"41c35e380758e6f1e2eadc037406e9f2933e80625681aba0dcfe25638fd3813c","last_reissued_at":"2026-07-05T07:44:00.357483Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:44:00.357483Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evading Data Contamination Detection for Language Models is (too) Easy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Jasper Dekoninck, Marc Fischer, Mark Niklas M\\\"uller, Martin Vechev, Maximilian Baader","submitted_at":"2024-02-05T09:10:32Z","abstract_excerpt":"Large language models are widespread, with their performance on benchmarks frequently guiding user preferences for one model over another. However, the vast amount of data these models are trained on can inadvertently lead to contamination with public benchmarks, thus compromising performance measurements. While recently developed contamination detection methods try to address this issue, they overlook the possibility of deliberate contamination by malicious model providers aiming to evade detection. We argue that this setting is of crucial importance as it casts doubt on the reliability of pu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.02823","kind":"arxiv","version":2},"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/2402.02823/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":"2402.02823","created_at":"2026-07-05T07:44:00.357543+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.02823v2","created_at":"2026-07-05T07:44:00.357543+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.02823","created_at":"2026-07-05T07:44:00.357543+00:00"},{"alias_kind":"pith_short_12","alias_value":"IHBV4OAHLDTP","created_at":"2026-07-05T07:44:00.357543+00:00"},{"alias_kind":"pith_short_16","alias_value":"IHBV4OAHLDTPDYXK","created_at":"2026-07-05T07:44:00.357543+00:00"},{"alias_kind":"pith_short_8","alias_value":"IHBV4OAH","created_at":"2026-07-05T07:44:00.357543+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26133","citing_title":"Pretraining Data Exposure in Large Language Models: A Survey of Membership Inference, Data Contamination, and Security Implications","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24079","citing_title":"TRACER: A Semantic-Aware Framework for Fine-Grained Contamination Detection in Code LLMs","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2406.04244","citing_title":"Benchmark Data Contamination of Large Language Models: A Survey","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21856","citing_title":"The Illusion of Reasoning: Exposing Evasive Data Contamination in LLMs via Zero-CoT Truncation","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2508.05452","citing_title":"LLMEval-Fair: A Large-Scale Longitudinal Study on Robust and Fair Evaluation of Large Language Models","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K","json":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K.json","graph_json":"https://pith.science/api/pith-number/IHBV4OAHLDTPDYXK3QBXIBXJ6K/graph.json","events_json":"https://pith.science/api/pith-number/IHBV4OAHLDTPDYXK3QBXIBXJ6K/events.json","paper":"https://pith.science/paper/IHBV4OAH"},"agent_actions":{"view_html":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K","download_json":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K.json","view_paper":"https://pith.science/paper/IHBV4OAH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.02823&json=true","fetch_graph":"https://pith.science/api/pith-number/IHBV4OAHLDTPDYXK3QBXIBXJ6K/graph.json","fetch_events":"https://pith.science/api/pith-number/IHBV4OAHLDTPDYXK3QBXIBXJ6K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K/action/storage_attestation","attest_author":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K/action/author_attestation","sign_citation":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K/action/citation_signature","submit_replication":"https://pith.science/pith/IHBV4OAHLDTPDYXK3QBXIBXJ6K/action/replication_record"}},"created_at":"2026-07-05T07:44:00.357543+00:00","updated_at":"2026-07-05T07:44:00.357543+00:00"}