{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KBVDFN5QIFUWCCKRLN36ZROBRD","short_pith_number":"pith:KBVDFN5Q","schema_version":"1.0","canonical_sha256":"506a32b7b041696109515b77ecc5c188eb0de3ad691bd476f49a63d236cca1f9","source":{"kind":"arxiv","id":"2605.15053","version":1},"attestation_state":"computed","paper":{"title":"TFGN: Task-Free, Replay-Free Continual Pre-Training Without Catastrophic Forgetting at LLM Scale","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Anurup Ganguli","submitted_at":"2026-05-14T16:46:26Z","abstract_excerpt":"Continually pre-training a large language model on heterogeneous text domains, without replay or task labels, has remained an unsolved architectural problem at LLM scale. Existing methods rely on replay buffers, task identifiers, regularization penalties that scale poorly, or sentence-classification-scale evaluation.\n  We introduce TFGN, an architectural overlay for transformer language models that produces input-conditioned, parameter-efficient updates while leaving the rest of the transformer unchanged. On six heterogeneous text domains (Prose, Python, Math, Biomedical, Chinese, JavaScript) "},"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":"2605.15053","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T16:46:26Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ea116ec5ee880a739d3230409ddf937660666bf91e6438614a1993dea2da79b3","abstract_canon_sha256":"2b54313f538f8fd7d43050c491eb0c45c47a06d1d483fdfbd79d8c7d093f7bad"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:54.369554Z","signature_b64":"rXUcZ9l/dXO80vuGvKobVT+aT4exGeCmTr1138vlIGNGgT6q6KKwy8BnYLHhkygulJXYRzaiRHq+wa4pTF9dDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"506a32b7b041696109515b77ecc5c188eb0de3ad691bd476f49a63d236cca1f9","last_reissued_at":"2026-05-17T23:38:54.368863Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:54.368863Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TFGN: Task-Free, Replay-Free Continual Pre-Training Without Catastrophic Forgetting at LLM Scale","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Anurup Ganguli","submitted_at":"2026-05-14T16:46:26Z","abstract_excerpt":"Continually pre-training a large language model on heterogeneous text domains, without replay or task labels, has remained an unsolved architectural problem at LLM scale. Existing methods rely on replay buffers, task identifiers, regularization penalties that scale poorly, or sentence-classification-scale evaluation.\n  We introduce TFGN, an architectural overlay for transformer language models that produces input-conditioned, parameter-efficient updates while leaving the rest of the transformer unchanged. On six heterogeneous text domains (Prose, Python, Math, Biomedical, Chinese, JavaScript) "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.15053","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":""},"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":"2605.15053","created_at":"2026-05-17T23:38:54.368967+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15053v1","created_at":"2026-05-17T23:38:54.368967+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15053","created_at":"2026-05-17T23:38:54.368967+00:00"},{"alias_kind":"pith_short_12","alias_value":"KBVDFN5QIFUW","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"KBVDFN5QIFUWCCKR","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"KBVDFN5Q","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD","json":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD.json","graph_json":"https://pith.science/api/pith-number/KBVDFN5QIFUWCCKRLN36ZROBRD/graph.json","events_json":"https://pith.science/api/pith-number/KBVDFN5QIFUWCCKRLN36ZROBRD/events.json","paper":"https://pith.science/paper/KBVDFN5Q"},"agent_actions":{"view_html":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD","download_json":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD.json","view_paper":"https://pith.science/paper/KBVDFN5Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15053&json=true","fetch_graph":"https://pith.science/api/pith-number/KBVDFN5QIFUWCCKRLN36ZROBRD/graph.json","fetch_events":"https://pith.science/api/pith-number/KBVDFN5QIFUWCCKRLN36ZROBRD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD/action/storage_attestation","attest_author":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD/action/author_attestation","sign_citation":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD/action/citation_signature","submit_replication":"https://pith.science/pith/KBVDFN5QIFUWCCKRLN36ZROBRD/action/replication_record"}},"created_at":"2026-05-17T23:38:54.368967+00:00","updated_at":"2026-05-17T23:38:54.368967+00:00"}