{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QVD4L32Z2F5ALHWMTX4IZCJSKQ","short_pith_number":"pith:QVD4L32Z","schema_version":"1.0","canonical_sha256":"8547c5ef59d17a059ecc9df88c8932541eb817378dfcf055f3946ab3bf6f7b87","source":{"kind":"arxiv","id":"2606.24998","version":1},"attestation_state":"computed","paper":{"title":"Internal Data Repetition Destroys Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bo He, David Donoho, Jessica Chudnovsky, Joshua Kazdan, Mehmet Donmez, Noam Levi, Rylan Schaeffer, Sanmi Koyejo, Yegor Denisov-Blanch","submitted_at":"2026-06-23T16:02:40Z","abstract_excerpt":"Language models are running out of high-quality training data, and even aggressively deduplicated corpora retain some amount of repetition. Earlier controlled studies predated Chinchilla-style scaling laws and could only measure the cost of repetition indirectly. We revisit repetition in the Chinchilla era, using a fitted no-repetition scaling law to report Compute-Equivalent Gain and Compute-Equivalent Loss. We show that under this modernized paradigm, repetition damage is systematic in three ways. First, holding compute allocated to repeated data constant, eval loss peaks at an intermediate "},"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":"2606.24998","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-23T16:02:40Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f315da6f562fb1d59436e1186c94297f650d9bd8b235c1536e56ee49241d251c","abstract_canon_sha256":"01946e2fa0e7047c2ae67b60a0c81192c32dc99b4036e8586f173f4018aec15e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-25T00:17:47.875772Z","signature_b64":"QkLNXLQrz4ZW/9Caj7AzVynF9S7x8Up6vchOH9FKhJtzveVtaptWod+lnEDPRtc7D232uiojTkKUMncvgO0ZDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8547c5ef59d17a059ecc9df88c8932541eb817378dfcf055f3946ab3bf6f7b87","last_reissued_at":"2026-06-25T00:17:47.875367Z","signature_status":"signed_v1","first_computed_at":"2026-06-25T00:17:47.875367Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Internal Data Repetition Destroys Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bo He, David Donoho, Jessica Chudnovsky, Joshua Kazdan, Mehmet Donmez, Noam Levi, Rylan Schaeffer, Sanmi Koyejo, Yegor Denisov-Blanch","submitted_at":"2026-06-23T16:02:40Z","abstract_excerpt":"Language models are running out of high-quality training data, and even aggressively deduplicated corpora retain some amount of repetition. Earlier controlled studies predated Chinchilla-style scaling laws and could only measure the cost of repetition indirectly. We revisit repetition in the Chinchilla era, using a fitted no-repetition scaling law to report Compute-Equivalent Gain and Compute-Equivalent Loss. We show that under this modernized paradigm, repetition damage is systematic in three ways. First, holding compute allocated to repeated data constant, eval loss peaks at an intermediate "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.24998","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/2606.24998/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":"2606.24998","created_at":"2026-06-25T00:17:47.875457+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.24998v1","created_at":"2026-06-25T00:17:47.875457+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.24998","created_at":"2026-06-25T00:17:47.875457+00:00"},{"alias_kind":"pith_short_12","alias_value":"QVD4L32Z2F5A","created_at":"2026-06-25T00:17:47.875457+00:00"},{"alias_kind":"pith_short_16","alias_value":"QVD4L32Z2F5ALHWM","created_at":"2026-06-25T00:17:47.875457+00:00"},{"alias_kind":"pith_short_8","alias_value":"QVD4L32Z","created_at":"2026-06-25T00:17:47.875457+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/QVD4L32Z2F5ALHWMTX4IZCJSKQ","json":"https://pith.science/pith/QVD4L32Z2F5ALHWMTX4IZCJSKQ.json","graph_json":"https://pith.science/api/pith-number/QVD4L32Z2F5ALHWMTX4IZCJSKQ/graph.json","events_json":"https://pith.science/api/pith-number/QVD4L32Z2F5ALHWMTX4IZCJSKQ/events.json","paper":"https://pith.science/paper/QVD4L32Z"},"agent_actions":{"view_html":"https://pith.science/pith/QVD4L32Z2F5ALHWMTX4IZCJSKQ","download_json":"https://pith.science/pith/QVD4L32Z2F5ALHWMTX4IZCJSKQ.json","view_paper":"https://pith.science/paper/QVD4L32Z","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.24998&json=true","fetch_graph":"https://pith.science/api/pith-number/QVD4L32Z2F5ALHWMTX4IZCJSKQ/graph.json","fetch_events":"https://pith.science/api/pith-number/QVD4L32Z2F5ALHWMTX4IZCJSKQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QVD4L32Z2F5ALHWMTX4IZCJSKQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QVD4L32Z2F5ALHWMTX4IZCJSKQ/action/storage_attestation","attest_author":"https://pith.science/pith/QVD4L32Z2F5ALHWMTX4IZCJSKQ/action/author_attestation","sign_citation":"https://pith.science/pith/QVD4L32Z2F5ALHWMTX4IZCJSKQ/action/citation_signature","submit_replication":"https://pith.science/pith/QVD4L32Z2F5ALHWMTX4IZCJSKQ/action/replication_record"}},"created_at":"2026-06-25T00:17:47.875457+00:00","updated_at":"2026-06-25T00:17:47.875457+00:00"}