{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZZW3P6TZIJFOYAMBEMK2RUZWWN","short_pith_number":"pith:ZZW3P6TZ","schema_version":"1.0","canonical_sha256":"ce6db7fa79424aec01812315a8d336b354eedf9d14e9aa23823261de35e72464","source":{"kind":"arxiv","id":"2407.15892","version":4},"attestation_state":"computed","paper":{"title":"Mini-Sequence Transformer: Optimizing Intermediate Memory for Long Sequences Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Anima Anandkumar, Beidi Chen, Cheng Luo, Jiawei Zhao, Zhuoming Chen","submitted_at":"2024-07-22T01:52:30Z","abstract_excerpt":"We introduce Mini-Sequence Transformer (MsT), a simple and effective methodology for highly efficient and accurate LLM training with extremely long sequences. MsT partitions input sequences and iteratively processes mini-sequences to reduce intermediate memory usage. Integrated with activation recomputation, it enables significant memory savings in both forward and backward passes. In experiments with the Llama3-8B model, with MsT, we measure no degradation in throughput or convergence even with 12x longer sequences than standard implementations. MsT is fully general, implementation-agnostic, "},"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":"2407.15892","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-07-22T01:52:30Z","cross_cats_sorted":[],"title_canon_sha256":"1db49b857107fb17de483083d8687226fe96bcbcbd26549049d5af732ec05183","abstract_canon_sha256":"4e54422611b62e4e2b79a2a3ce935c54393b2fdaad30fc947dcecc1f289de915"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:33:35.550784Z","signature_b64":"cYM1qgPOIMbIW6ydosRSkH4SKP01+a6Q3VvEroLqQ4EKR252VceeziiffHh/aYl7wLQ5Obk1QWhVEzHbasWQAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ce6db7fa79424aec01812315a8d336b354eedf9d14e9aa23823261de35e72464","last_reissued_at":"2026-07-05T09:33:35.550288Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:33:35.550288Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mini-Sequence Transformer: Optimizing Intermediate Memory for Long Sequences Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Anima Anandkumar, Beidi Chen, Cheng Luo, Jiawei Zhao, Zhuoming Chen","submitted_at":"2024-07-22T01:52:30Z","abstract_excerpt":"We introduce Mini-Sequence Transformer (MsT), a simple and effective methodology for highly efficient and accurate LLM training with extremely long sequences. MsT partitions input sequences and iteratively processes mini-sequences to reduce intermediate memory usage. Integrated with activation recomputation, it enables significant memory savings in both forward and backward passes. In experiments with the Llama3-8B model, with MsT, we measure no degradation in throughput or convergence even with 12x longer sequences than standard implementations. MsT is fully general, implementation-agnostic, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.15892","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/2407.15892/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":"2407.15892","created_at":"2026-07-05T09:33:35.550356+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.15892v4","created_at":"2026-07-05T09:33:35.550356+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.15892","created_at":"2026-07-05T09:33:35.550356+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZZW3P6TZIJFO","created_at":"2026-07-05T09:33:35.550356+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZZW3P6TZIJFOYAMB","created_at":"2026-07-05T09:33:35.550356+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZZW3P6TZ","created_at":"2026-07-05T09:33:35.550356+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.03077","citing_title":"StreamBP: Memory-Efficient Exact Backpropagation for Long Sequence Training of LLMs","ref_index":20,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN","json":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN.json","graph_json":"https://pith.science/api/pith-number/ZZW3P6TZIJFOYAMBEMK2RUZWWN/graph.json","events_json":"https://pith.science/api/pith-number/ZZW3P6TZIJFOYAMBEMK2RUZWWN/events.json","paper":"https://pith.science/paper/ZZW3P6TZ"},"agent_actions":{"view_html":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN","download_json":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN.json","view_paper":"https://pith.science/paper/ZZW3P6TZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.15892&json=true","fetch_graph":"https://pith.science/api/pith-number/ZZW3P6TZIJFOYAMBEMK2RUZWWN/graph.json","fetch_events":"https://pith.science/api/pith-number/ZZW3P6TZIJFOYAMBEMK2RUZWWN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN/action/storage_attestation","attest_author":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN/action/author_attestation","sign_citation":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN/action/citation_signature","submit_replication":"https://pith.science/pith/ZZW3P6TZIJFOYAMBEMK2RUZWWN/action/replication_record"}},"created_at":"2026-07-05T09:33:35.550356+00:00","updated_at":"2026-07-05T09:33:35.550356+00:00"}