{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NB6NEJ2ELIO4R7C4INR6LOCLU5","short_pith_number":"pith:NB6NEJ2E","schema_version":"1.0","canonical_sha256":"687cd227445a1dc8fc5c4363e5b84ba7668ffae009063d4564d6d7e00e5f19c4","source":{"kind":"arxiv","id":"2405.13226","version":2},"attestation_state":"computed","paper":{"title":"Dataset Decomposition: Faster LLM Training with Variable Sequence Length Curriculum","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Cem Koc, Chun-Liang Li, Hadi Pouransari, Jen-Hao Rick Chang, Oncel Tuzel, Pavan Kumar Anasosalu Vasu, Vaishaal Shankar","submitted_at":"2024-05-21T22:26:01Z","abstract_excerpt":"Large language models (LLMs) are commonly trained on datasets consisting of fixed-length token sequences. These datasets are created by randomly concatenating documents of various lengths and then chunking them into sequences of a predetermined target length (concat-and-chunk). Recent attention implementations mask cross-document attention, reducing the effective length of a chunk of tokens. Additionally, training on long sequences becomes computationally prohibitive due to the quadratic cost of attention. In this study, we introduce dataset decomposition, a novel variable sequence length trai"},"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":"2405.13226","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-21T22:26:01Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b24ae81f10ba21b6c75ec72e267f6ba6342afa16eb9eec605f26732bad3620e4","abstract_canon_sha256":"c8a84a69cd26d8a8b6ce53af3bc18f47be8d7456b10e69b668f5f31a23662c49"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:57:42.833727Z","signature_b64":"UBbNAgyMAwGTtPWGQcnFkvPGPRzbFq+qVd7KvfHnt+Kgkc7eLlFvIhC93I2wdPuOGM8NJQhFBq/MT+VhHK7vBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"687cd227445a1dc8fc5c4363e5b84ba7668ffae009063d4564d6d7e00e5f19c4","last_reissued_at":"2026-07-05T09:57:42.833219Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:57:42.833219Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dataset Decomposition: Faster LLM Training with Variable Sequence Length Curriculum","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Cem Koc, Chun-Liang Li, Hadi Pouransari, Jen-Hao Rick Chang, Oncel Tuzel, Pavan Kumar Anasosalu Vasu, Vaishaal Shankar","submitted_at":"2024-05-21T22:26:01Z","abstract_excerpt":"Large language models (LLMs) are commonly trained on datasets consisting of fixed-length token sequences. These datasets are created by randomly concatenating documents of various lengths and then chunking them into sequences of a predetermined target length (concat-and-chunk). Recent attention implementations mask cross-document attention, reducing the effective length of a chunk of tokens. Additionally, training on long sequences becomes computationally prohibitive due to the quadratic cost of attention. In this study, we introduce dataset decomposition, a novel variable sequence length trai"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.13226","kind":"arxiv","version":2},"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/2405.13226/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":"2405.13226","created_at":"2026-07-05T09:57:42.833269+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.13226v2","created_at":"2026-07-05T09:57:42.833269+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.13226","created_at":"2026-07-05T09:57:42.833269+00:00"},{"alias_kind":"pith_short_12","alias_value":"NB6NEJ2ELIO4","created_at":"2026-07-05T09:57:42.833269+00:00"},{"alias_kind":"pith_short_16","alias_value":"NB6NEJ2ELIO4R7C4","created_at":"2026-07-05T09:57:42.833269+00:00"},{"alias_kind":"pith_short_8","alias_value":"NB6NEJ2E","created_at":"2026-07-05T09:57:42.833269+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2406.11794","citing_title":"DataComp-LM: In search of the next generation of training sets for language models","ref_index":144,"is_internal_anchor":false},{"citing_arxiv_id":"2405.21060","citing_title":"Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality","ref_index":78,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5","json":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5.json","graph_json":"https://pith.science/api/pith-number/NB6NEJ2ELIO4R7C4INR6LOCLU5/graph.json","events_json":"https://pith.science/api/pith-number/NB6NEJ2ELIO4R7C4INR6LOCLU5/events.json","paper":"https://pith.science/paper/NB6NEJ2E"},"agent_actions":{"view_html":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5","download_json":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5.json","view_paper":"https://pith.science/paper/NB6NEJ2E","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.13226&json=true","fetch_graph":"https://pith.science/api/pith-number/NB6NEJ2ELIO4R7C4INR6LOCLU5/graph.json","fetch_events":"https://pith.science/api/pith-number/NB6NEJ2ELIO4R7C4INR6LOCLU5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5/action/storage_attestation","attest_author":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5/action/author_attestation","sign_citation":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5/action/citation_signature","submit_replication":"https://pith.science/pith/NB6NEJ2ELIO4R7C4INR6LOCLU5/action/replication_record"}},"created_at":"2026-07-05T09:57:42.833269+00:00","updated_at":"2026-07-05T09:57:42.833269+00:00"}