{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:H4TLQIROHFPUW7CL5JMLGSPCEQ","short_pith_number":"pith:H4TLQIRO","schema_version":"1.0","canonical_sha256":"3f26b8222e395f4b7c4bea58b349e224399ccace92fac4ed13c59b097bd7cc4e","source":{"kind":"arxiv","id":"2502.03589","version":1},"attestation_state":"computed","paper":{"title":"HACK: Homomorphic Acceleration via Compression of the Key-Value Cache for Disaggregated LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Haiying Shen, Michael Mitzenmacher, Minlan Yu, Ran Ben Basat, Shay Vargaftik, Zeyu Zhang","submitted_at":"2025-02-05T20:09:51Z","abstract_excerpt":"Disaggregated Large Language Model (LLM) inference has gained popularity as it separates the computation-intensive prefill stage from the memory-intensive decode stage, avoiding the prefill-decode interference and improving resource utilization. However, transmitting Key-Value (KV) data between the two stages can be a bottleneck, especially for long prompts. Additionally, the computation time overhead for prefill and decode is key for optimizing Job Completion Time (JCT), and KV data size can become prohibitive for long prompts and sequences. Existing KV quantization methods can alleviate the "},"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":"2502.03589","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2025-02-05T20:09:51Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"3f76ecd3250a375d90bce8f5814901dcaf4c49a9b233ecb3bcbb14495adb8025","abstract_canon_sha256":"e249aac166f2522862b738c3629fb70115d7e37a0bfa8036cc57a79d508e0f65"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:10:12.556531Z","signature_b64":"y5PkhYQuBaE1o5QXy8AIOy6rXBF7nKDdonVS8FDzxcX/nZfzONrzsn9lyiqqYaJrq0zXXj2rSZ4gtyasB5yyDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3f26b8222e395f4b7c4bea58b349e224399ccace92fac4ed13c59b097bd7cc4e","last_reissued_at":"2026-07-05T10:10:12.556068Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:10:12.556068Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HACK: Homomorphic Acceleration via Compression of the Key-Value Cache for Disaggregated LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Haiying Shen, Michael Mitzenmacher, Minlan Yu, Ran Ben Basat, Shay Vargaftik, Zeyu Zhang","submitted_at":"2025-02-05T20:09:51Z","abstract_excerpt":"Disaggregated Large Language Model (LLM) inference has gained popularity as it separates the computation-intensive prefill stage from the memory-intensive decode stage, avoiding the prefill-decode interference and improving resource utilization. However, transmitting Key-Value (KV) data between the two stages can be a bottleneck, especially for long prompts. Additionally, the computation time overhead for prefill and decode is key for optimizing Job Completion Time (JCT), and KV data size can become prohibitive for long prompts and sequences. Existing KV quantization methods can alleviate the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.03589","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/2502.03589/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":"2502.03589","created_at":"2026-07-05T10:10:12.556127+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.03589v1","created_at":"2026-07-05T10:10:12.556127+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.03589","created_at":"2026-07-05T10:10:12.556127+00:00"},{"alias_kind":"pith_short_12","alias_value":"H4TLQIROHFPU","created_at":"2026-07-05T10:10:12.556127+00:00"},{"alias_kind":"pith_short_16","alias_value":"H4TLQIROHFPUW7CL","created_at":"2026-07-05T10:10:12.556127+00:00"},{"alias_kind":"pith_short_8","alias_value":"H4TLQIRO","created_at":"2026-07-05T10:10:12.556127+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/H4TLQIROHFPUW7CL5JMLGSPCEQ","json":"https://pith.science/pith/H4TLQIROHFPUW7CL5JMLGSPCEQ.json","graph_json":"https://pith.science/api/pith-number/H4TLQIROHFPUW7CL5JMLGSPCEQ/graph.json","events_json":"https://pith.science/api/pith-number/H4TLQIROHFPUW7CL5JMLGSPCEQ/events.json","paper":"https://pith.science/paper/H4TLQIRO"},"agent_actions":{"view_html":"https://pith.science/pith/H4TLQIROHFPUW7CL5JMLGSPCEQ","download_json":"https://pith.science/pith/H4TLQIROHFPUW7CL5JMLGSPCEQ.json","view_paper":"https://pith.science/paper/H4TLQIRO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.03589&json=true","fetch_graph":"https://pith.science/api/pith-number/H4TLQIROHFPUW7CL5JMLGSPCEQ/graph.json","fetch_events":"https://pith.science/api/pith-number/H4TLQIROHFPUW7CL5JMLGSPCEQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/H4TLQIROHFPUW7CL5JMLGSPCEQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/H4TLQIROHFPUW7CL5JMLGSPCEQ/action/storage_attestation","attest_author":"https://pith.science/pith/H4TLQIROHFPUW7CL5JMLGSPCEQ/action/author_attestation","sign_citation":"https://pith.science/pith/H4TLQIROHFPUW7CL5JMLGSPCEQ/action/citation_signature","submit_replication":"https://pith.science/pith/H4TLQIROHFPUW7CL5JMLGSPCEQ/action/replication_record"}},"created_at":"2026-07-05T10:10:12.556127+00:00","updated_at":"2026-07-05T10:10:12.556127+00:00"}