{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CU347JJNSNVJEWLSN5467RZVQE","short_pith_number":"pith:CU347JJN","schema_version":"1.0","canonical_sha256":"1537cfa52d936a9259726f79efc7358108614c57ac71fc25670ba09cf562dd99","source":{"kind":"arxiv","id":"2407.12835","version":2},"attestation_state":"computed","paper":{"title":"Regurgitative Training: The Value of Real Data in Training Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.CL","authors_text":"Dandan Qiao, Jinghui Zhang, Mochen Yang, Qiang Wei","submitted_at":"2024-07-03T18:42:55Z","abstract_excerpt":"What happens if we train a new Large Language Model (LLM) using data that are at least partially generated by other LLMs? The explosive success of LLMs means that a substantial amount of content online will be generated by LLMs rather than humans, which will inevitably enter the training datasets of next-generation LLMs. We evaluate the implications of such \"regurgitative training\" on LLM performance. Through fine-tuning GPT-3.5 with data generated either by itself or by other LLMs in a machine translation task, we find strong evidence that regurgitative training clearly handicaps the performa"},"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":"2407.12835","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-03T18:42:55Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"81ef85a9fc02d39a1291d9c5624dc13a552fba028d2e5bdab98515c2a4b24ffb","abstract_canon_sha256":"c5678275dab2612117be3f8a0f1e5bb098a4a8bbf351bf5ad74391d1760e57c9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:48:32.448170Z","signature_b64":"xq+qzf5sfQlExKHI97F256KF0vCa4GaRcuXeJ20fUNFKiSXAkXvaB0Yt1jT1OhAOi4Evi8CthlkkNxw3ls1OCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1537cfa52d936a9259726f79efc7358108614c57ac71fc25670ba09cf562dd99","last_reissued_at":"2026-07-05T08:48:32.447693Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:48:32.447693Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Regurgitative Training: The Value of Real Data in Training Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.CL","authors_text":"Dandan Qiao, Jinghui Zhang, Mochen Yang, Qiang Wei","submitted_at":"2024-07-03T18:42:55Z","abstract_excerpt":"What happens if we train a new Large Language Model (LLM) using data that are at least partially generated by other LLMs? The explosive success of LLMs means that a substantial amount of content online will be generated by LLMs rather than humans, which will inevitably enter the training datasets of next-generation LLMs. We evaluate the implications of such \"regurgitative training\" on LLM performance. Through fine-tuning GPT-3.5 with data generated either by itself or by other LLMs in a machine translation task, we find strong evidence that regurgitative training clearly handicaps the performa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.12835","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/2407.12835/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":"2407.12835","created_at":"2026-07-05T08:48:32.447749+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.12835v2","created_at":"2026-07-05T08:48:32.447749+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.12835","created_at":"2026-07-05T08:48:32.447749+00:00"},{"alias_kind":"pith_short_12","alias_value":"CU347JJNSNVJ","created_at":"2026-07-05T08:48:32.447749+00:00"},{"alias_kind":"pith_short_16","alias_value":"CU347JJNSNVJEWLS","created_at":"2026-07-05T08:48:32.447749+00:00"},{"alias_kind":"pith_short_8","alias_value":"CU347JJN","created_at":"2026-07-05T08:48:32.447749+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2505.11336","citing_title":"XtraGPT: Context-Aware and Controllable Academic Paper Revision via Human-AI Collaboration","ref_index":86,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE","json":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE.json","graph_json":"https://pith.science/api/pith-number/CU347JJNSNVJEWLSN5467RZVQE/graph.json","events_json":"https://pith.science/api/pith-number/CU347JJNSNVJEWLSN5467RZVQE/events.json","paper":"https://pith.science/paper/CU347JJN"},"agent_actions":{"view_html":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE","download_json":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE.json","view_paper":"https://pith.science/paper/CU347JJN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.12835&json=true","fetch_graph":"https://pith.science/api/pith-number/CU347JJNSNVJEWLSN5467RZVQE/graph.json","fetch_events":"https://pith.science/api/pith-number/CU347JJNSNVJEWLSN5467RZVQE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE/action/storage_attestation","attest_author":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE/action/author_attestation","sign_citation":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE/action/citation_signature","submit_replication":"https://pith.science/pith/CU347JJNSNVJEWLSN5467RZVQE/action/replication_record"}},"created_at":"2026-07-05T08:48:32.447749+00:00","updated_at":"2026-07-05T08:48:32.447749+00:00"}