{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:HQ3PBZTHLGCDXP7J3E7FEQQUJ5","short_pith_number":"pith:HQ3PBZTH","schema_version":"1.0","canonical_sha256":"3c36f0e66759843bbfe9d93e5242144f6f88feff35db31bcf21d13c0a56c9864","source":{"kind":"arxiv","id":"2302.03241","version":4},"attestation_state":"computed","paper":{"title":"Continual Pre-training of Language Models","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.NE"],"primary_cat":"cs.CL","authors_text":"Bing Liu, Gyuhak Kim, Haowei Lin, Tatsuya Konishi, Yijia Shao, Zixuan Ke","submitted_at":"2023-02-07T03:57:55Z","abstract_excerpt":"Language models (LMs) have been instrumental for the rapid advance of natural language processing. This paper studies continual pre-training of LMs, in particular, continual domain-adaptive pre-training (or continual DAP-training). Existing research has shown that further pre-training an LM using a domain corpus to adapt the LM to the domain can improve the end-task performance in the domain. This paper proposes a novel method to continually DAP-train an LM with a sequence of unlabeled domain corpora to adapt the LM to these domains to improve their end-task performances. The key novelty of ou"},"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":"2302.03241","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.CL","submitted_at":"2023-02-07T03:57:55Z","cross_cats_sorted":["cs.AI","cs.LG","cs.NE"],"title_canon_sha256":"ac2099373c87aa214c0d3d5946a61f56ed92c947a1e50850f49c59a6b12e79a0","abstract_canon_sha256":"dabcdd5b87e0f1c6ca5e50ce2c3791fcf8b134541d445e9683b792c36a034658"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:00:16.237818Z","signature_b64":"4QQbfikuK5agOIPvRDpbK13K6yr2jCYGNxcSiblENR/+7L0AK/fNAvBmPi4x8P2reaMCx7P5Ip13ikn3YW1MCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3c36f0e66759843bbfe9d93e5242144f6f88feff35db31bcf21d13c0a56c9864","last_reissued_at":"2026-07-05T06:00:16.237293Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:00:16.237293Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Continual Pre-training of Language Models","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.NE"],"primary_cat":"cs.CL","authors_text":"Bing Liu, Gyuhak Kim, Haowei Lin, Tatsuya Konishi, Yijia Shao, Zixuan Ke","submitted_at":"2023-02-07T03:57:55Z","abstract_excerpt":"Language models (LMs) have been instrumental for the rapid advance of natural language processing. This paper studies continual pre-training of LMs, in particular, continual domain-adaptive pre-training (or continual DAP-training). Existing research has shown that further pre-training an LM using a domain corpus to adapt the LM to the domain can improve the end-task performance in the domain. This paper proposes a novel method to continually DAP-train an LM with a sequence of unlabeled domain corpora to adapt the LM to these domains to improve their end-task performances. The key novelty of ou"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.03241","kind":"arxiv","version":4},"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/2302.03241/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":"2302.03241","created_at":"2026-07-05T06:00:16.237362+00:00"},{"alias_kind":"arxiv_version","alias_value":"2302.03241v4","created_at":"2026-07-05T06:00:16.237362+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.03241","created_at":"2026-07-05T06:00:16.237362+00:00"},{"alias_kind":"pith_short_12","alias_value":"HQ3PBZTHLGCD","created_at":"2026-07-05T06:00:16.237362+00:00"},{"alias_kind":"pith_short_16","alias_value":"HQ3PBZTHLGCDXP7J","created_at":"2026-07-05T06:00:16.237362+00:00"},{"alias_kind":"pith_short_8","alias_value":"HQ3PBZTH","created_at":"2026-07-05T06:00:16.237362+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26097","citing_title":"Forgetting in Language Models: Capacity, Optimization, and Self-Generated Replay","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2409.04777","citing_title":"Optimization Hyper-parameter Laws for Large Language Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15156","citing_title":"MeMo: Memory as a Model","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20296","citing_title":"Spectral Unforgetting: Post-Hoc Recovery of Damaged Capabilities Without Retraining","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2509.07177","citing_title":"Towards EnergyGPT: A Large Language Model Specialized for the Energy Sector","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2509.17183","citing_title":"LifeAlign: Lifelong Alignment for Large Language Models with Memory-Augmented Focalized Preference Optimization","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2601.21577","citing_title":"Collaborative Parameter Learning: Mitigating Forgetting via Parameter-Level Gradient Analysis","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2601.16175","citing_title":"Learning to Discover at Test Time","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15156","citing_title":"MeMo: Memory as a Model","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10640","citing_title":"Towards Understanding Continual Factual Knowledge Acquisition of Language Models: From Theory to Algorithm","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05776","citing_title":"HEDP: A Hybrid Energy-Distance Prompt-based Framework for Domain Incremental Learning","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5","json":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5.json","graph_json":"https://pith.science/api/pith-number/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/graph.json","events_json":"https://pith.science/api/pith-number/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/events.json","paper":"https://pith.science/paper/HQ3PBZTH"},"agent_actions":{"view_html":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5","download_json":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5.json","view_paper":"https://pith.science/paper/HQ3PBZTH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2302.03241&json=true","fetch_graph":"https://pith.science/api/pith-number/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/graph.json","fetch_events":"https://pith.science/api/pith-number/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/action/storage_attestation","attest_author":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/action/author_attestation","sign_citation":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/action/citation_signature","submit_replication":"https://pith.science/pith/HQ3PBZTHLGCDXP7J3E7FEQQUJ5/action/replication_record"}},"created_at":"2026-07-05T06:00:16.237362+00:00","updated_at":"2026-07-05T06:00:16.237362+00:00"}