{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FMIIDHTLIGAQIWDZZZ3B5VE2XL","short_pith_number":"pith:FMIIDHTL","schema_version":"1.0","canonical_sha256":"2b10819e6b4181045879ce761ed49abad7e2e88b2c0275383ec23d49cea7665f","source":{"kind":"arxiv","id":"2501.14743","version":1},"attestation_state":"computed","paper":{"title":"KVDirect: Distributed Disaggregated LLM Inference","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG","cs.PF"],"primary_cat":"cs.DC","authors_text":"Chenyu Jiang, Dezhi YU, Fanlong Meng, Hang Liu, Jinlai Xu, Mengyuan Chao, Rain Jiang, Shiyang Chen, Wei Xu","submitted_at":"2024-12-13T21:54:16Z","abstract_excerpt":"Large Language Models (LLMs) have become the new foundation for many applications, reshaping human society like a storm. Disaggregated inference, which separates prefill and decode stages, is a promising approach to improving hardware utilization and service quality. However, due to inefficient inter-node communication, existing systems restrict disaggregated inference to a single node, limiting resource allocation flexibility and reducing service capacity. This paper introduces KVDirect, which optimizes KV cache transfer to enable a distributed disaggregated LLM inference. KVDirect achieves t"},"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":"2501.14743","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.DC","submitted_at":"2024-12-13T21:54:16Z","cross_cats_sorted":["cs.LG","cs.PF"],"title_canon_sha256":"d740e037425338474500fe89177870e4f78a5b3c61720bc854dc05e18ff6f398","abstract_canon_sha256":"d53ca952a30c66d7f9a6237337175e8080436e5f4d04c94f37ae2058fa577bc2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:05:17.590252Z","signature_b64":"WvdxBLONghfEAUQ1OhS3HJy5olc9Pd3QpmLOs0MSYbuYdRt6MRf5F6NSblWW8w2wehJtZt97icS5ywIhF0EnDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2b10819e6b4181045879ce761ed49abad7e2e88b2c0275383ec23d49cea7665f","last_reissued_at":"2026-07-05T10:05:17.589784Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:05:17.589784Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"KVDirect: Distributed Disaggregated LLM Inference","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG","cs.PF"],"primary_cat":"cs.DC","authors_text":"Chenyu Jiang, Dezhi YU, Fanlong Meng, Hang Liu, Jinlai Xu, Mengyuan Chao, Rain Jiang, Shiyang Chen, Wei Xu","submitted_at":"2024-12-13T21:54:16Z","abstract_excerpt":"Large Language Models (LLMs) have become the new foundation for many applications, reshaping human society like a storm. Disaggregated inference, which separates prefill and decode stages, is a promising approach to improving hardware utilization and service quality. However, due to inefficient inter-node communication, existing systems restrict disaggregated inference to a single node, limiting resource allocation flexibility and reducing service capacity. This paper introduces KVDirect, which optimizes KV cache transfer to enable a distributed disaggregated LLM inference. KVDirect achieves t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.14743","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/2501.14743/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":"2501.14743","created_at":"2026-07-05T10:05:17.589840+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.14743v1","created_at":"2026-07-05T10:05:17.589840+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.14743","created_at":"2026-07-05T10:05:17.589840+00:00"},{"alias_kind":"pith_short_12","alias_value":"FMIIDHTLIGAQ","created_at":"2026-07-05T10:05:17.589840+00:00"},{"alias_kind":"pith_short_16","alias_value":"FMIIDHTLIGAQIWDZ","created_at":"2026-07-05T10:05:17.589840+00:00"},{"alias_kind":"pith_short_8","alias_value":"FMIIDHTL","created_at":"2026-07-05T10:05:17.589840+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19746","citing_title":"SAC: Disaggregated KV Cache System for Sparse Attention LLMs with CXL","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06113","citing_title":"Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06113","citing_title":"Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL","json":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL.json","graph_json":"https://pith.science/api/pith-number/FMIIDHTLIGAQIWDZZZ3B5VE2XL/graph.json","events_json":"https://pith.science/api/pith-number/FMIIDHTLIGAQIWDZZZ3B5VE2XL/events.json","paper":"https://pith.science/paper/FMIIDHTL"},"agent_actions":{"view_html":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL","download_json":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL.json","view_paper":"https://pith.science/paper/FMIIDHTL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.14743&json=true","fetch_graph":"https://pith.science/api/pith-number/FMIIDHTLIGAQIWDZZZ3B5VE2XL/graph.json","fetch_events":"https://pith.science/api/pith-number/FMIIDHTLIGAQIWDZZZ3B5VE2XL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL/action/storage_attestation","attest_author":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL/action/author_attestation","sign_citation":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL/action/citation_signature","submit_replication":"https://pith.science/pith/FMIIDHTLIGAQIWDZZZ3B5VE2XL/action/replication_record"}},"created_at":"2026-07-05T10:05:17.589840+00:00","updated_at":"2026-07-05T10:05:17.589840+00:00"}