{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ICUNQHOTV2A2NJPDO744CJ2T4I","short_pith_number":"pith:ICUNQHOT","schema_version":"1.0","canonical_sha256":"40a8d81dd3ae81a6a5e377f9c12753e2287d1f14837a9724e9243b388098574e","source":{"kind":"arxiv","id":"2409.07431","version":2},"attestation_state":"computed","paper":{"title":"Synthetic continued pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Emmanuel Cand\\`es, Neil Band, Shuangping Li, Tatsunori Hashimoto, Zitong Yang","submitted_at":"2024-09-11T17:21:59Z","abstract_excerpt":"Pretraining on large-scale, unstructured internet text enables language models to acquire a significant amount of world knowledge. However, this knowledge acquisition is data-inefficient--to learn a given fact, models must be trained on hundreds to thousands of diverse representations of it. This poses a challenge when adapting a pretrained model to a small corpus of domain-specific documents, where each fact may appear rarely or only once. We propose to bridge this gap with synthetic continued pretraining: using the small domain-specific corpus to synthesize a large corpus more amenable to le"},"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":"2409.07431","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-09-11T17:21:59Z","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"title_canon_sha256":"c99c16dd4f7e777b0fe2bd85c7aaacf66d24e58c9c8e8bc0d341b0be37f98959","abstract_canon_sha256":"f672d8560a494efddfeed06c3010fceac91edd21ea763f5643b18c730ba05400"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:15:14.417148Z","signature_b64":"3Ek5+EsB1imUt7UrBBzt4DT5yFE1ttTxn6Gk/wmN4hISZNv7cPyjKxH5jp/Kc7SUA09RrlU894S2clH2kmsPDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"40a8d81dd3ae81a6a5e377f9c12753e2287d1f14837a9724e9243b388098574e","last_reissued_at":"2026-07-05T09:15:14.416683Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:15:14.416683Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Synthetic continued pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Emmanuel Cand\\`es, Neil Band, Shuangping Li, Tatsunori Hashimoto, Zitong Yang","submitted_at":"2024-09-11T17:21:59Z","abstract_excerpt":"Pretraining on large-scale, unstructured internet text enables language models to acquire a significant amount of world knowledge. However, this knowledge acquisition is data-inefficient--to learn a given fact, models must be trained on hundreds to thousands of diverse representations of it. This poses a challenge when adapting a pretrained model to a small corpus of domain-specific documents, where each fact may appear rarely or only once. We propose to bridge this gap with synthetic continued pretraining: using the small domain-specific corpus to synthesize a large corpus more amenable to le"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.07431","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/2409.07431/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":"2409.07431","created_at":"2026-07-05T09:15:14.416741+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.07431v2","created_at":"2026-07-05T09:15:14.416741+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.07431","created_at":"2026-07-05T09:15:14.416741+00:00"},{"alias_kind":"pith_short_12","alias_value":"ICUNQHOTV2A2","created_at":"2026-07-05T09:15:14.416741+00:00"},{"alias_kind":"pith_short_16","alias_value":"ICUNQHOTV2A2NJPD","created_at":"2026-07-05T09:15:14.416741+00:00"},{"alias_kind":"pith_short_8","alias_value":"ICUNQHOT","created_at":"2026-07-05T09:15:14.416741+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01849","citing_title":"ContinuousBench: Can Differentially Private Synthetic Text Improve Capabilities?","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2606.32002","citing_title":"Self-Study Reconsidered: The Hidden Fragility of Learning from Self-Generated QA","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20189","citing_title":"SOLAR: A Self-Optimizing Open-Ended Autonomous Agent for Lifelong Learning and Continual Adaptation","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19394","citing_title":"EmbGen: Teaching with Reassembled Corpora","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2602.12705","citing_title":"MedXIAOHE: A Comprehensive Recipe for Building Medical MLLMs","ref_index":64,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I","json":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I.json","graph_json":"https://pith.science/api/pith-number/ICUNQHOTV2A2NJPDO744CJ2T4I/graph.json","events_json":"https://pith.science/api/pith-number/ICUNQHOTV2A2NJPDO744CJ2T4I/events.json","paper":"https://pith.science/paper/ICUNQHOT"},"agent_actions":{"view_html":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I","download_json":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I.json","view_paper":"https://pith.science/paper/ICUNQHOT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.07431&json=true","fetch_graph":"https://pith.science/api/pith-number/ICUNQHOTV2A2NJPDO744CJ2T4I/graph.json","fetch_events":"https://pith.science/api/pith-number/ICUNQHOTV2A2NJPDO744CJ2T4I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I/action/storage_attestation","attest_author":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I/action/author_attestation","sign_citation":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I/action/citation_signature","submit_replication":"https://pith.science/pith/ICUNQHOTV2A2NJPDO744CJ2T4I/action/replication_record"}},"created_at":"2026-07-05T09:15:14.416741+00:00","updated_at":"2026-07-05T09:15:14.416741+00:00"}