{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:3CD4KJ2TPNF6CIDIQECBSTRHFJ","short_pith_number":"pith:3CD4KJ2T","schema_version":"1.0","canonical_sha256":"d887c527537b4be120688104194e272a7d0c536c14c88cbadabd28801f276160","source":{"kind":"arxiv","id":"2410.00161","version":2},"attestation_state":"computed","paper":{"title":"KV-Compress: Paged KV-Cache Compression with Variable Compression Rates per Attention Head","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Isaac Rehg","submitted_at":"2024-09-30T19:09:13Z","abstract_excerpt":"Context lengths of Large Language Models (LLMs) have exploded in recent years, with 128k-token context becoming a standard and million-token context becoming a reality. Efficiently supporting long-context inference remains challenging as the memory that must be allocated in key-value (KV) cache for a generation scales with its context length, limiting the number of long-context requests that can be served concurrently under a given memory budget. KV cache compression can mitigate this issue by removing under-utilized KVs from each attention head's cache and reducing its memory footprint. Highe"},"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":"2410.00161","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-09-30T19:09:13Z","cross_cats_sorted":[],"title_canon_sha256":"0a9ace2844c16e004362f0e067b6ac0265317481313a386f1ed17566c6403c0d","abstract_canon_sha256":"b7b4461cb4b89dd367bf5e225c262e96487533676820203eec725fb30cf87960"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:16:53.254877Z","signature_b64":"JRSDFtckQMCCm9RKmDwU20WT3pd/4etkJzGp2XxWOQFkEjT2rdcWlA8TiN4eZgDE6TSKLMBMF2EZkwvRqAUGDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d887c527537b4be120688104194e272a7d0c536c14c88cbadabd28801f276160","last_reissued_at":"2026-07-05T09:16:53.254381Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:16:53.254381Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"KV-Compress: Paged KV-Cache Compression with Variable Compression Rates per Attention Head","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Isaac Rehg","submitted_at":"2024-09-30T19:09:13Z","abstract_excerpt":"Context lengths of Large Language Models (LLMs) have exploded in recent years, with 128k-token context becoming a standard and million-token context becoming a reality. Efficiently supporting long-context inference remains challenging as the memory that must be allocated in key-value (KV) cache for a generation scales with its context length, limiting the number of long-context requests that can be served concurrently under a given memory budget. KV cache compression can mitigate this issue by removing under-utilized KVs from each attention head's cache and reducing its memory footprint. Highe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.00161","kind":"arxiv","version":2},"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/2410.00161/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":"2410.00161","created_at":"2026-07-05T09:16:53.254438+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.00161v2","created_at":"2026-07-05T09:16:53.254438+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.00161","created_at":"2026-07-05T09:16:53.254438+00:00"},{"alias_kind":"pith_short_12","alias_value":"3CD4KJ2TPNF6","created_at":"2026-07-05T09:16:53.254438+00:00"},{"alias_kind":"pith_short_16","alias_value":"3CD4KJ2TPNF6CIDI","created_at":"2026-07-05T09:16:53.254438+00:00"},{"alias_kind":"pith_short_8","alias_value":"3CD4KJ2T","created_at":"2026-07-05T09:16:53.254438+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.29563","citing_title":"Coverage-Driven KV Cache Eviction for Efficient and Improved Inference of LLM","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23200","citing_title":"Adaptive Mass-Segmented KV Compression for Long-Context Reasoning","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ","json":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ.json","graph_json":"https://pith.science/api/pith-number/3CD4KJ2TPNF6CIDIQECBSTRHFJ/graph.json","events_json":"https://pith.science/api/pith-number/3CD4KJ2TPNF6CIDIQECBSTRHFJ/events.json","paper":"https://pith.science/paper/3CD4KJ2T"},"agent_actions":{"view_html":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ","download_json":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ.json","view_paper":"https://pith.science/paper/3CD4KJ2T","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.00161&json=true","fetch_graph":"https://pith.science/api/pith-number/3CD4KJ2TPNF6CIDIQECBSTRHFJ/graph.json","fetch_events":"https://pith.science/api/pith-number/3CD4KJ2TPNF6CIDIQECBSTRHFJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ/action/storage_attestation","attest_author":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ/action/author_attestation","sign_citation":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ/action/citation_signature","submit_replication":"https://pith.science/pith/3CD4KJ2TPNF6CIDIQECBSTRHFJ/action/replication_record"}},"created_at":"2026-07-05T09:16:53.254438+00:00","updated_at":"2026-07-05T09:16:53.254438+00:00"}