{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FP53QKYREJJIVY5CPLSD3TG5HR","short_pith_number":"pith:FP53QKYR","schema_version":"1.0","canonical_sha256":"2bfbb82b1122528ae3a27ae43dccdd3c7994bdb6cd69b0fd42a5bda51dfedf4a","source":{"kind":"arxiv","id":"2411.15871","version":1},"attestation_state":"computed","paper":{"title":"Hiding Communication Cost in Distributed LLM Training via Micro-batch Co-execution","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Chaoyi Ruan, Chengjie Tang, Cheng Li, HaiQuan Wang, Jia He, Jiaqi Ruan, Xiaosong Ma","submitted_at":"2024-11-24T15:15:46Z","abstract_excerpt":"The growth of Large Language Models (LLMs) has necessitated large-scale distributed training. Highly optimized frameworks, however, still suffer significant losses in Model FLOPS utilization (often below 50%) due to large communication volumes. Meanwhile, our comprehensive profiling shows that the computation- and communication-intensive operators overlap well.\n  This paper introduces DHelix, a novel micro-structure that dramatically improves the efficiency of LLM training inspired by the DNA structure. Central to DHelix's design is Strand Interleaving (SI), which views the continuous stream o"},"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":"2411.15871","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2024-11-24T15:15:46Z","cross_cats_sorted":[],"title_canon_sha256":"70cd83b9e3267baa2deffcea1b6bb5d16b156408538b21aac496255b64dc3212","abstract_canon_sha256":"7827fa1c25ad04d10cebadd7e387e78d5f2f852c45884bf1dc185516c65df078"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:39:42.544189Z","signature_b64":"rVxnuhYoxyapprh17FFCCwj23xjCbXEa8Y4IDwxH6LC8psnQRTxk3wnQ0IWyLfYEAto9VyZhRD1cnAbYpjW3Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2bfbb82b1122528ae3a27ae43dccdd3c7994bdb6cd69b0fd42a5bda51dfedf4a","last_reissued_at":"2026-07-05T09:39:42.543672Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:39:42.543672Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hiding Communication Cost in Distributed LLM Training via Micro-batch Co-execution","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Chaoyi Ruan, Chengjie Tang, Cheng Li, HaiQuan Wang, Jia He, Jiaqi Ruan, Xiaosong Ma","submitted_at":"2024-11-24T15:15:46Z","abstract_excerpt":"The growth of Large Language Models (LLMs) has necessitated large-scale distributed training. Highly optimized frameworks, however, still suffer significant losses in Model FLOPS utilization (often below 50%) due to large communication volumes. Meanwhile, our comprehensive profiling shows that the computation- and communication-intensive operators overlap well.\n  This paper introduces DHelix, a novel micro-structure that dramatically improves the efficiency of LLM training inspired by the DNA structure. Central to DHelix's design is Strand Interleaving (SI), which views the continuous stream o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.15871","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/2411.15871/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":"2411.15871","created_at":"2026-07-05T09:39:42.543732+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.15871v1","created_at":"2026-07-05T09:39:42.543732+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.15871","created_at":"2026-07-05T09:39:42.543732+00:00"},{"alias_kind":"pith_short_12","alias_value":"FP53QKYREJJI","created_at":"2026-07-05T09:39:42.543732+00:00"},{"alias_kind":"pith_short_16","alias_value":"FP53QKYREJJIVY5C","created_at":"2026-07-05T09:39:42.543732+00:00"},{"alias_kind":"pith_short_8","alias_value":"FP53QKYR","created_at":"2026-07-05T09:39:42.543732+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.24088","citing_title":"TACO: Efficient Communication Compression of Intermediate Tensors for Scalable Tensor-Parallel LLM Training","ref_index":54,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR","json":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR.json","graph_json":"https://pith.science/api/pith-number/FP53QKYREJJIVY5CPLSD3TG5HR/graph.json","events_json":"https://pith.science/api/pith-number/FP53QKYREJJIVY5CPLSD3TG5HR/events.json","paper":"https://pith.science/paper/FP53QKYR"},"agent_actions":{"view_html":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR","download_json":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR.json","view_paper":"https://pith.science/paper/FP53QKYR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.15871&json=true","fetch_graph":"https://pith.science/api/pith-number/FP53QKYREJJIVY5CPLSD3TG5HR/graph.json","fetch_events":"https://pith.science/api/pith-number/FP53QKYREJJIVY5CPLSD3TG5HR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR/action/storage_attestation","attest_author":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR/action/author_attestation","sign_citation":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR/action/citation_signature","submit_replication":"https://pith.science/pith/FP53QKYREJJIVY5CPLSD3TG5HR/action/replication_record"}},"created_at":"2026-07-05T09:39:42.543732+00:00","updated_at":"2026-07-05T09:39:42.543732+00:00"}