{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RB763DXV7MR7DQSHKAX73AEARM","short_pith_number":"pith:RB763DXV","schema_version":"1.0","canonical_sha256":"887fed8ef5fb23f1c247502ffd80808b073b549ca142aa9cec571458cd930124","source":{"kind":"arxiv","id":"2506.22033","version":1},"attestation_state":"computed","paper":{"title":"SiPipe: Bridging the CPU-GPU Utilization Gap for Efficient Pipeline-Parallel LLM Inference","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Bohan Zhao, Yongchao He, Zheng Cao","submitted_at":"2025-06-27T09:27:04Z","abstract_excerpt":"As inference workloads for large language models (LLMs) scale to meet growing user demand, pipeline parallelism (PP) has become a widely adopted strategy for multi-GPU deployment, particularly in cross-node setups, to improve key-value (KV) cache capacity and inference throughput. However, PP suffers from inherent inefficiencies caused by three types of execution bubbles-load-imbalance, intra-stage, and inter-stage-which limit pipeline saturation. We present SiPipe, a heterogeneous pipeline design that improves throughput by leveraging underutilized CPU resources to offload auxiliary computati"},"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":"2506.22033","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.DC","submitted_at":"2025-06-27T09:27:04Z","cross_cats_sorted":[],"title_canon_sha256":"36803eb871ef2cc017577ec62ac39951845562dbe7625e864771cad61bc66f75","abstract_canon_sha256":"e07b2db0f88e8f24bb9268b02825b3a173ad8ef351279b1b18d2e7871a40a52e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:28:20.815791Z","signature_b64":"ZqUIRMeshNounLIryY0D3f5Hco31XwB58zd5buzdj8SWfMl44fPtmwlqlcXICgUSYyojNm2Piz61ESybhUU6Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"887fed8ef5fb23f1c247502ffd80808b073b549ca142aa9cec571458cd930124","last_reissued_at":"2026-07-05T11:28:20.815294Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:28:20.815294Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SiPipe: Bridging the CPU-GPU Utilization Gap for Efficient Pipeline-Parallel LLM Inference","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Bohan Zhao, Yongchao He, Zheng Cao","submitted_at":"2025-06-27T09:27:04Z","abstract_excerpt":"As inference workloads for large language models (LLMs) scale to meet growing user demand, pipeline parallelism (PP) has become a widely adopted strategy for multi-GPU deployment, particularly in cross-node setups, to improve key-value (KV) cache capacity and inference throughput. However, PP suffers from inherent inefficiencies caused by three types of execution bubbles-load-imbalance, intra-stage, and inter-stage-which limit pipeline saturation. We present SiPipe, a heterogeneous pipeline design that improves throughput by leveraging underutilized CPU resources to offload auxiliary computati"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.22033","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/2506.22033/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":"2506.22033","created_at":"2026-07-05T11:28:20.815363+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.22033v1","created_at":"2026-07-05T11:28:20.815363+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.22033","created_at":"2026-07-05T11:28:20.815363+00:00"},{"alias_kind":"pith_short_12","alias_value":"RB763DXV7MR7","created_at":"2026-07-05T11:28:20.815363+00:00"},{"alias_kind":"pith_short_16","alias_value":"RB763DXV7MR7DQSH","created_at":"2026-07-05T11:28:20.815363+00:00"},{"alias_kind":"pith_short_8","alias_value":"RB763DXV","created_at":"2026-07-05T11:28:20.815363+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/RB763DXV7MR7DQSHKAX73AEARM","json":"https://pith.science/pith/RB763DXV7MR7DQSHKAX73AEARM.json","graph_json":"https://pith.science/api/pith-number/RB763DXV7MR7DQSHKAX73AEARM/graph.json","events_json":"https://pith.science/api/pith-number/RB763DXV7MR7DQSHKAX73AEARM/events.json","paper":"https://pith.science/paper/RB763DXV"},"agent_actions":{"view_html":"https://pith.science/pith/RB763DXV7MR7DQSHKAX73AEARM","download_json":"https://pith.science/pith/RB763DXV7MR7DQSHKAX73AEARM.json","view_paper":"https://pith.science/paper/RB763DXV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.22033&json=true","fetch_graph":"https://pith.science/api/pith-number/RB763DXV7MR7DQSHKAX73AEARM/graph.json","fetch_events":"https://pith.science/api/pith-number/RB763DXV7MR7DQSHKAX73AEARM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RB763DXV7MR7DQSHKAX73AEARM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RB763DXV7MR7DQSHKAX73AEARM/action/storage_attestation","attest_author":"https://pith.science/pith/RB763DXV7MR7DQSHKAX73AEARM/action/author_attestation","sign_citation":"https://pith.science/pith/RB763DXV7MR7DQSHKAX73AEARM/action/citation_signature","submit_replication":"https://pith.science/pith/RB763DXV7MR7DQSHKAX73AEARM/action/replication_record"}},"created_at":"2026-07-05T11:28:20.815363+00:00","updated_at":"2026-07-05T11:28:20.815363+00:00"}