{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FHIBFT2FCAH6TNYKWKJEBOYM4L","short_pith_number":"pith:FHIBFT2F","schema_version":"1.0","canonical_sha256":"29d012cf45100fe9b70ab29240bb0ce2faaef089f12e26f51d99bb3f725d3bbb","source":{"kind":"arxiv","id":"2502.21231","version":1},"attestation_state":"computed","paper":{"title":"ByteScale: Efficient Scaling of LLM Training with a 2048K Context Length on More Than 12,000 GPUs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Bin Cui, Fangcheng Fu, Haibin Lin, Hao Ge, Junda Feng, Lei Zuo, Qi Huang, Xiaonan Nie, Xin Liu","submitted_at":"2025-02-28T17:01:03Z","abstract_excerpt":"Scaling long-context ability is essential for Large Language Models (LLMs). To amortize the memory consumption across multiple devices in long-context training, inter-data partitioning (a.k.a. Data Parallelism) and intra-data partitioning (a.k.a. Context Parallelism) are commonly used. Current training frameworks predominantly treat the two techniques as orthogonal, and establish static communication groups to organize the devices as a static mesh (e.g., a 2D mesh). However, the sequences for LLM training typically vary in lengths, no matter for texts, multi-modalities or reinforcement learnin"},"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":"2502.21231","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-02-28T17:01:03Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"2e6a45ba58ae11bf412d31c8074d2163d792aa1908613db288ddd2b69bf975c7","abstract_canon_sha256":"f3dc7f4a96c544f4991bcbb28fd7751712d541c30c7db47dc874fb443251a96e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:01:21.108283Z","signature_b64":"6P5+kD62as8iHIoO6G0cpo6KAtxsXu6zvLI0FRZNR3pY7PweHvEZV/rwEbqDFHjFCNqAgIiIGjFRNJQDmM1xAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"29d012cf45100fe9b70ab29240bb0ce2faaef089f12e26f51d99bb3f725d3bbb","last_reissued_at":"2026-07-05T12:01:21.107866Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:01:21.107866Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ByteScale: Efficient Scaling of LLM Training with a 2048K Context Length on More Than 12,000 GPUs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Bin Cui, Fangcheng Fu, Haibin Lin, Hao Ge, Junda Feng, Lei Zuo, Qi Huang, Xiaonan Nie, Xin Liu","submitted_at":"2025-02-28T17:01:03Z","abstract_excerpt":"Scaling long-context ability is essential for Large Language Models (LLMs). To amortize the memory consumption across multiple devices in long-context training, inter-data partitioning (a.k.a. Data Parallelism) and intra-data partitioning (a.k.a. Context Parallelism) are commonly used. Current training frameworks predominantly treat the two techniques as orthogonal, and establish static communication groups to organize the devices as a static mesh (e.g., a 2D mesh). However, the sequences for LLM training typically vary in lengths, no matter for texts, multi-modalities or reinforcement learnin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.21231","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/2502.21231/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":"2502.21231","created_at":"2026-07-05T12:01:21.107925+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.21231v1","created_at":"2026-07-05T12:01:21.107925+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.21231","created_at":"2026-07-05T12:01:21.107925+00:00"},{"alias_kind":"pith_short_12","alias_value":"FHIBFT2FCAH6","created_at":"2026-07-05T12:01:21.107925+00:00"},{"alias_kind":"pith_short_16","alias_value":"FHIBFT2FCAH6TNYK","created_at":"2026-07-05T12:01:21.107925+00:00"},{"alias_kind":"pith_short_8","alias_value":"FHIBFT2F","created_at":"2026-07-05T12:01:21.107925+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08476","citing_title":"FlashCP: Load-Balanced Communication-Efficient Context Parallelism for LLM Training","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2504.09844","citing_title":"MegaScale-Data: Scaling Dataloader for Multisource Large Foundation Model Training","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2510.18830","citing_title":"MTraining: Distributed Dynamic Sparse Attention for Efficient Ultra-Long Context Training","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2509.21275","citing_title":"InfiniPipe: Elastic Pipeline Parallelism for Efficient Variable-Length Long-Context LLM Training","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2505.13211","citing_title":"MAGI-1: Autoregressive Video Generation at Scale","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08962","citing_title":"MegaScale-Omni: A Hyper-Scale, Workload-Resilient System for MultiModal LLM Training in Production","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2602.15763","citing_title":"GLM-5: from Vibe Coding to Agentic Engineering","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21026","citing_title":"MCAP: Deployment-Time Layer Profiling for Memory-Constrained LLM Inference","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L","json":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L.json","graph_json":"https://pith.science/api/pith-number/FHIBFT2FCAH6TNYKWKJEBOYM4L/graph.json","events_json":"https://pith.science/api/pith-number/FHIBFT2FCAH6TNYKWKJEBOYM4L/events.json","paper":"https://pith.science/paper/FHIBFT2F"},"agent_actions":{"view_html":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L","download_json":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L.json","view_paper":"https://pith.science/paper/FHIBFT2F","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.21231&json=true","fetch_graph":"https://pith.science/api/pith-number/FHIBFT2FCAH6TNYKWKJEBOYM4L/graph.json","fetch_events":"https://pith.science/api/pith-number/FHIBFT2FCAH6TNYKWKJEBOYM4L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L/action/storage_attestation","attest_author":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L/action/author_attestation","sign_citation":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L/action/citation_signature","submit_replication":"https://pith.science/pith/FHIBFT2FCAH6TNYKWKJEBOYM4L/action/replication_record"}},"created_at":"2026-07-05T12:01:21.107925+00:00","updated_at":"2026-07-05T12:01:21.107925+00:00"}