{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:4BKKBVA7SXOLXEOOSQ7N3FFE75","short_pith_number":"pith:4BKKBVA7","schema_version":"1.0","canonical_sha256":"e054a0d41f95dcbb91ce943edd94a4ff6c830333a3a2fac96197adbc30024284","source":{"kind":"arxiv","id":"2305.12281","version":1},"attestation_state":"computed","paper":{"title":"Lifelong Language Pretraining with Distribution-Specialized Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Claire Cu, James Laudon, Nan Du, Wuyang Chen, Yanping Huang, Yanqi Zhou, Zhifeng Chen","submitted_at":"2023-05-20T21:15:19Z","abstract_excerpt":"Pretraining on a large-scale corpus has become a standard method to build general language models (LMs). Adapting a model to new data distributions targeting different downstream tasks poses significant challenges. Naive fine-tuning may incur catastrophic forgetting when the over-parameterized LMs overfit the new data but fail to preserve the pretrained features. Lifelong learning (LLL) aims to enable information systems to learn from a continuous data stream across time. However, most prior work modifies the training recipe assuming a static fixed network architecture. We find that additional"},"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":"2305.12281","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-20T21:15:19Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"e852540e385b6bbcaa02d57136b98157b2833933c4104dee13674babf1f875ba","abstract_canon_sha256":"6338d151c0ce4c415b01fc1655dbd532d63c168c6fbb0ea9217fd1321de528f1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:12:08.817783Z","signature_b64":"l6jvv2v/rNL3oYLsuHv6gG/g/149U4ZgUu/XPETkmKzg4Tx4ZUbe96hJWkKruA9mZMvziMvkPL0PsQfZ/SqkCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e054a0d41f95dcbb91ce943edd94a4ff6c830333a3a2fac96197adbc30024284","last_reissued_at":"2026-07-05T06:12:08.817306Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:12:08.817306Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Lifelong Language Pretraining with Distribution-Specialized Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Claire Cu, James Laudon, Nan Du, Wuyang Chen, Yanping Huang, Yanqi Zhou, Zhifeng Chen","submitted_at":"2023-05-20T21:15:19Z","abstract_excerpt":"Pretraining on a large-scale corpus has become a standard method to build general language models (LMs). Adapting a model to new data distributions targeting different downstream tasks poses significant challenges. Naive fine-tuning may incur catastrophic forgetting when the over-parameterized LMs overfit the new data but fail to preserve the pretrained features. Lifelong learning (LLL) aims to enable information systems to learn from a continuous data stream across time. However, most prior work modifies the training recipe assuming a static fixed network architecture. We find that additional"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.12281","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/2305.12281/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":"2305.12281","created_at":"2026-07-05T06:12:08.817375+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.12281v1","created_at":"2026-07-05T06:12:08.817375+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.12281","created_at":"2026-07-05T06:12:08.817375+00:00"},{"alias_kind":"pith_short_12","alias_value":"4BKKBVA7SXOL","created_at":"2026-07-05T06:12:08.817375+00:00"},{"alias_kind":"pith_short_16","alias_value":"4BKKBVA7SXOLXEOO","created_at":"2026-07-05T06:12:08.817375+00:00"},{"alias_kind":"pith_short_8","alias_value":"4BKKBVA7","created_at":"2026-07-05T06:12:08.817375+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.15053","citing_title":"TFGN: Task-Free, Replay-Free Continual Pre-Training Without Catastrophic Forgetting at LLM Scale","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75","json":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75.json","graph_json":"https://pith.science/api/pith-number/4BKKBVA7SXOLXEOOSQ7N3FFE75/graph.json","events_json":"https://pith.science/api/pith-number/4BKKBVA7SXOLXEOOSQ7N3FFE75/events.json","paper":"https://pith.science/paper/4BKKBVA7"},"agent_actions":{"view_html":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75","download_json":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75.json","view_paper":"https://pith.science/paper/4BKKBVA7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.12281&json=true","fetch_graph":"https://pith.science/api/pith-number/4BKKBVA7SXOLXEOOSQ7N3FFE75/graph.json","fetch_events":"https://pith.science/api/pith-number/4BKKBVA7SXOLXEOOSQ7N3FFE75/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75/action/storage_attestation","attest_author":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75/action/author_attestation","sign_citation":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75/action/citation_signature","submit_replication":"https://pith.science/pith/4BKKBVA7SXOLXEOOSQ7N3FFE75/action/replication_record"}},"created_at":"2026-07-05T06:12:08.817375+00:00","updated_at":"2026-07-05T06:12:08.817375+00:00"}