{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QRAORAQWGHCQXWUIRSLTW33GJX","short_pith_number":"pith:QRAORAQW","schema_version":"1.0","canonical_sha256":"8440e8821631c50bda888c973b6f664dd63dcf0dcad050030445b37b90733c62","source":{"kind":"arxiv","id":"2605.26099","version":1},"attestation_state":"computed","paper":{"title":"Language Models Need Sleep","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Giulia Fanti, Sangyun Lee, Sean McLeish, Tom Goldstein","submitted_at":"2026-05-25T17:55:39Z","abstract_excerpt":"Transformer-based large language models are increasingly used for long-horizon tasks; however, their attention mechanism scales poorly with context length. To handle this, we study a sleep-like consolidation mechanism in which a model periodically converts recent context into persistent fast weights before clearing its key-value cache. During sleep, the model performs $N$ offline recurrent passes over the accumulated context and updates the fast weights in its state-space model (SSM) blocks through a learned local rule. During inference, this shifts extra computation to sleep while preserving "},"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":"2605.26099","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-25T17:55:39Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"94675ebb39ced2f3a389608d8eb8d2901c6346a45affa2134e052a338f404606","abstract_canon_sha256":"79fc4195a781947395bef37f585dd4736d4f395ab38929cc7207d72b2717f367"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:05:27.416100Z","signature_b64":"fXAnbteljJ0QWEsZhH3g99qrbX5Bog0mHEVo3UWChig6p3FLWn+9dS8vgJbuFeY9Xu9IJqEyRLzRAFBCaECIAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8440e8821631c50bda888c973b6f664dd63dcf0dcad050030445b37b90733c62","last_reissued_at":"2026-05-26T02:05:27.415581Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:05:27.415581Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Language Models Need Sleep","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Giulia Fanti, Sangyun Lee, Sean McLeish, Tom Goldstein","submitted_at":"2026-05-25T17:55:39Z","abstract_excerpt":"Transformer-based large language models are increasingly used for long-horizon tasks; however, their attention mechanism scales poorly with context length. To handle this, we study a sleep-like consolidation mechanism in which a model periodically converts recent context into persistent fast weights before clearing its key-value cache. During sleep, the model performs $N$ offline recurrent passes over the accumulated context and updates the fast weights in its state-space model (SSM) blocks through a learned local rule. During inference, this shifts extra computation to sleep while preserving "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.26099","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/2605.26099/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":"2605.26099","created_at":"2026-05-26T02:05:27.415675+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.26099v1","created_at":"2026-05-26T02:05:27.415675+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.26099","created_at":"2026-05-26T02:05:27.415675+00:00"},{"alias_kind":"pith_short_12","alias_value":"QRAORAQWGHCQ","created_at":"2026-05-26T02:05:27.415675+00:00"},{"alias_kind":"pith_short_16","alias_value":"QRAORAQWGHCQXWUI","created_at":"2026-05-26T02:05:27.415675+00:00"},{"alias_kind":"pith_short_8","alias_value":"QRAORAQW","created_at":"2026-05-26T02:05:27.415675+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/QRAORAQWGHCQXWUIRSLTW33GJX","json":"https://pith.science/pith/QRAORAQWGHCQXWUIRSLTW33GJX.json","graph_json":"https://pith.science/api/pith-number/QRAORAQWGHCQXWUIRSLTW33GJX/graph.json","events_json":"https://pith.science/api/pith-number/QRAORAQWGHCQXWUIRSLTW33GJX/events.json","paper":"https://pith.science/paper/QRAORAQW"},"agent_actions":{"view_html":"https://pith.science/pith/QRAORAQWGHCQXWUIRSLTW33GJX","download_json":"https://pith.science/pith/QRAORAQWGHCQXWUIRSLTW33GJX.json","view_paper":"https://pith.science/paper/QRAORAQW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.26099&json=true","fetch_graph":"https://pith.science/api/pith-number/QRAORAQWGHCQXWUIRSLTW33GJX/graph.json","fetch_events":"https://pith.science/api/pith-number/QRAORAQWGHCQXWUIRSLTW33GJX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QRAORAQWGHCQXWUIRSLTW33GJX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QRAORAQWGHCQXWUIRSLTW33GJX/action/storage_attestation","attest_author":"https://pith.science/pith/QRAORAQWGHCQXWUIRSLTW33GJX/action/author_attestation","sign_citation":"https://pith.science/pith/QRAORAQWGHCQXWUIRSLTW33GJX/action/citation_signature","submit_replication":"https://pith.science/pith/QRAORAQWGHCQXWUIRSLTW33GJX/action/replication_record"}},"created_at":"2026-05-26T02:05:27.415675+00:00","updated_at":"2026-05-26T02:05:27.415675+00:00"}