{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:MKR7HBTMELECR4ELSIZOEOIOYU","short_pith_number":"pith:MKR7HBTM","schema_version":"1.0","canonical_sha256":"62a3f3866c22c828f08b9232e2390ec500648803ebea698dbd28040d1f66a4a5","source":{"kind":"arxiv","id":"2608.04074","version":1},"attestation_state":"computed","paper":{"title":"Spend Bits Where Queries Look: KV Cache Vector Quantization with Attention-Preserving Transforms","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.IT","eess.SP","math.IT"],"primary_cat":"cs.LG","authors_text":"Amir Ziashahabi, Antonio Ortega, Eduardo Pavez, Salman Avestimehr, Samuel Fern\\'andez-Mendui\\~na","submitted_at":"2026-08-04T16:10:59Z","abstract_excerpt":"Long-context LLM decoding reads the key-value (KV) cache at every step. Loading it takes longer than computing attention over it, so throughput is bandwidth-bound. Hence, reducing the cache size can raise both decoding speed and serving capacity. The challenge is to reduce cache size while preserving the attention products, keeping reconstruction cheap, and using a fixed per-token bit count. At two bits per element, the most competitive methods rely on orthogonal transforms. However, existing techniques are either data-oblivious or use the query statistics without deriving the transform from a"},"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":"2608.04074","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-04T16:10:59Z","cross_cats_sorted":["cs.AI","cs.IT","eess.SP","math.IT"],"title_canon_sha256":"e6929dab911e64d76760d843375ef4fa7686d1bc2d6422ef898c2d9c3094c83f","abstract_canon_sha256":"735bd29135ee7df0b77f4ad19d1ffb9ede70331bf65fd9b0b64b3bd6e8bb41dc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-06T01:28:39.576919Z","signature_b64":"qkL5v78Mm5TWK5nrQqs+c912L+eu4wmEAzaNXv23ccWT1XfoX0kS0RsYid9m5zPjwrIrHO8UQ/opeVcwcMboBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"62a3f3866c22c828f08b9232e2390ec500648803ebea698dbd28040d1f66a4a5","last_reissued_at":"2026-08-06T01:28:39.575406Z","signature_status":"signed_v1","first_computed_at":"2026-08-06T01:28:39.575406Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Spend Bits Where Queries Look: KV Cache Vector Quantization with Attention-Preserving Transforms","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.IT","eess.SP","math.IT"],"primary_cat":"cs.LG","authors_text":"Amir Ziashahabi, Antonio Ortega, Eduardo Pavez, Salman Avestimehr, Samuel Fern\\'andez-Mendui\\~na","submitted_at":"2026-08-04T16:10:59Z","abstract_excerpt":"Long-context LLM decoding reads the key-value (KV) cache at every step. Loading it takes longer than computing attention over it, so throughput is bandwidth-bound. Hence, reducing the cache size can raise both decoding speed and serving capacity. The challenge is to reduce cache size while preserving the attention products, keeping reconstruction cheap, and using a fixed per-token bit count. At two bits per element, the most competitive methods rely on orthogonal transforms. However, existing techniques are either data-oblivious or use the query statistics without deriving the transform from a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.04074","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/2608.04074/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":"2608.04074","created_at":"2026-08-06T01:28:39.577258+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.04074v1","created_at":"2026-08-06T01:28:39.577258+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.04074","created_at":"2026-08-06T01:28:39.577258+00:00"},{"alias_kind":"pith_short_12","alias_value":"MKR7HBTMELEC","created_at":"2026-08-06T01:28:39.577258+00:00"},{"alias_kind":"pith_short_16","alias_value":"MKR7HBTMELECR4EL","created_at":"2026-08-06T01:28:39.577258+00:00"},{"alias_kind":"pith_short_8","alias_value":"MKR7HBTM","created_at":"2026-08-06T01:28:39.577258+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/MKR7HBTMELECR4ELSIZOEOIOYU","json":"https://pith.science/pith/MKR7HBTMELECR4ELSIZOEOIOYU.json","graph_json":"https://pith.science/api/pith-number/MKR7HBTMELECR4ELSIZOEOIOYU/graph.json","events_json":"https://pith.science/api/pith-number/MKR7HBTMELECR4ELSIZOEOIOYU/events.json","paper":"https://pith.science/paper/MKR7HBTM"},"agent_actions":{"view_html":"https://pith.science/pith/MKR7HBTMELECR4ELSIZOEOIOYU","download_json":"https://pith.science/pith/MKR7HBTMELECR4ELSIZOEOIOYU.json","view_paper":"https://pith.science/paper/MKR7HBTM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.04074&json=true","fetch_graph":"https://pith.science/api/pith-number/MKR7HBTMELECR4ELSIZOEOIOYU/graph.json","fetch_events":"https://pith.science/api/pith-number/MKR7HBTMELECR4ELSIZOEOIOYU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MKR7HBTMELECR4ELSIZOEOIOYU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MKR7HBTMELECR4ELSIZOEOIOYU/action/storage_attestation","attest_author":"https://pith.science/pith/MKR7HBTMELECR4ELSIZOEOIOYU/action/author_attestation","sign_citation":"https://pith.science/pith/MKR7HBTMELECR4ELSIZOEOIOYU/action/citation_signature","submit_replication":"https://pith.science/pith/MKR7HBTMELECR4ELSIZOEOIOYU/action/replication_record"}},"created_at":"2026-08-06T01:28:39.577258+00:00","updated_at":"2026-08-06T01:28:39.577258+00:00"}