{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FESQ3MSVYUEOTJPP4JEE52WUCK","short_pith_number":"pith:FESQ3MSV","schema_version":"1.0","canonical_sha256":"29250db255c508e9a5efe2484eead4129e4153e6fc7b2a2337b795808715e08c","source":{"kind":"arxiv","id":"2504.05571","version":1},"attestation_state":"computed","paper":{"title":"Knowledge-Instruct: Effective Continual Pre-training from Limited Data using Instructions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Eitam Sheetrit, Meni Brief, Oded Ovadia, Rachel Lemberg","submitted_at":"2025-04-08T00:00:36Z","abstract_excerpt":"While Large Language Models (LLMs) acquire vast knowledge during pre-training, they often lack domain-specific, new, or niche information. Continual pre-training (CPT) attempts to address this gap but suffers from catastrophic forgetting and inefficiencies in low-data regimes. We introduce Knowledge-Instruct, a novel approach to efficiently inject knowledge from limited corpora through pure instruction-tuning. By generating information-dense synthetic instruction data, it effectively integrates new knowledge while preserving general reasoning and instruction-following abilities. Knowledge-Inst"},"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":"2504.05571","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-08T00:00:36Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"bd5071b038c919fa1a20cb6a3e3527b89388f9dba570013c05461307d2b7842d","abstract_canon_sha256":"4a5ccbab7e649dd0180206c7cc7f9b7c2c1dabe67f3db479cd51d4a792241a94"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:46:05.951087Z","signature_b64":"A0syGJMMsX05ZBKQoa201jIVwEL/oz9fwKP9zk60kmCUImQ7kVP2dSFu30HdC0jE5hPG4/mhE0RtsoHP2omfBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"29250db255c508e9a5efe2484eead4129e4153e6fc7b2a2337b795808715e08c","last_reissued_at":"2026-07-05T10:46:05.950590Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:46:05.950590Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Knowledge-Instruct: Effective Continual Pre-training from Limited Data using Instructions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Eitam Sheetrit, Meni Brief, Oded Ovadia, Rachel Lemberg","submitted_at":"2025-04-08T00:00:36Z","abstract_excerpt":"While Large Language Models (LLMs) acquire vast knowledge during pre-training, they often lack domain-specific, new, or niche information. Continual pre-training (CPT) attempts to address this gap but suffers from catastrophic forgetting and inefficiencies in low-data regimes. We introduce Knowledge-Instruct, a novel approach to efficiently inject knowledge from limited corpora through pure instruction-tuning. By generating information-dense synthetic instruction data, it effectively integrates new knowledge while preserving general reasoning and instruction-following abilities. Knowledge-Inst"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.05571","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/2504.05571/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":"2504.05571","created_at":"2026-07-05T10:46:05.950652+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.05571v1","created_at":"2026-07-05T10:46:05.950652+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.05571","created_at":"2026-07-05T10:46:05.950652+00:00"},{"alias_kind":"pith_short_12","alias_value":"FESQ3MSVYUEO","created_at":"2026-07-05T10:46:05.950652+00:00"},{"alias_kind":"pith_short_16","alias_value":"FESQ3MSVYUEOTJPP","created_at":"2026-07-05T10:46:05.950652+00:00"},{"alias_kind":"pith_short_8","alias_value":"FESQ3MSV","created_at":"2026-07-05T10:46:05.950652+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01849","citing_title":"ContinuousBench: Can Differentially Private Synthetic Text Improve Capabilities?","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19394","citing_title":"EmbGen: Teaching with Reassembled Corpora","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2510.09474","citing_title":"Multimodal Policy Internalization for Conversational Agents","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK","json":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK.json","graph_json":"https://pith.science/api/pith-number/FESQ3MSVYUEOTJPP4JEE52WUCK/graph.json","events_json":"https://pith.science/api/pith-number/FESQ3MSVYUEOTJPP4JEE52WUCK/events.json","paper":"https://pith.science/paper/FESQ3MSV"},"agent_actions":{"view_html":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK","download_json":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK.json","view_paper":"https://pith.science/paper/FESQ3MSV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.05571&json=true","fetch_graph":"https://pith.science/api/pith-number/FESQ3MSVYUEOTJPP4JEE52WUCK/graph.json","fetch_events":"https://pith.science/api/pith-number/FESQ3MSVYUEOTJPP4JEE52WUCK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK/action/storage_attestation","attest_author":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK/action/author_attestation","sign_citation":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK/action/citation_signature","submit_replication":"https://pith.science/pith/FESQ3MSVYUEOTJPP4JEE52WUCK/action/replication_record"}},"created_at":"2026-07-05T10:46:05.950652+00:00","updated_at":"2026-07-05T10:46:05.950652+00:00"}