{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:3OU3AZ7GHOPP6DMIAR5CKXO3K6","short_pith_number":"pith:3OU3AZ7G","schema_version":"1.0","canonical_sha256":"dba9b067e63b9eff0d88047a255ddb57b5efdb8305b11fc9c9413680e6bc23e7","source":{"kind":"arxiv","id":"2503.23294","version":1},"attestation_state":"computed","paper":{"title":"Cocktail: Chunk-Adaptive Mixed-Precision Quantization for Long-Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bin Zhang, Jianzong Wang, Jiguang Wan, Wei Tao, Xiaoyang Qu","submitted_at":"2025-03-30T03:20:34Z","abstract_excerpt":"Recently, large language models (LLMs) have been able to handle longer and longer contexts. However, a context that is too long may cause intolerant inference latency and GPU memory usage. Existing methods propose mixed-precision quantization to the key-value (KV) cache in LLMs based on token granularity, which is time-consuming in the search process and hardware inefficient during computation. This paper introduces a novel approach called Cocktail, which employs chunk-adaptive mixed-precision quantization to optimize the KV cache. Cocktail consists of two modules: chunk-level quantization sea"},"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":"2503.23294","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-03-30T03:20:34Z","cross_cats_sorted":[],"title_canon_sha256":"cae934ce80414b902e7088e6a51512a4b62a6a35a48799c3744f2a030cbc875b","abstract_canon_sha256":"42ff105610df7c855ef04aa8dd728dddd8c44d47bcd26cf660b0dceed63075a1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:41:33.899713Z","signature_b64":"H+YonMp0M6T3nPabIM6nga0EmzUui24PwLQ587o1Q+zvXtRq1+19FayiVR7QzjevDOhgr3Nh/mNqXqamh25QBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dba9b067e63b9eff0d88047a255ddb57b5efdb8305b11fc9c9413680e6bc23e7","last_reissued_at":"2026-07-05T10:41:33.899229Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:41:33.899229Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cocktail: Chunk-Adaptive Mixed-Precision Quantization for Long-Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bin Zhang, Jianzong Wang, Jiguang Wan, Wei Tao, Xiaoyang Qu","submitted_at":"2025-03-30T03:20:34Z","abstract_excerpt":"Recently, large language models (LLMs) have been able to handle longer and longer contexts. However, a context that is too long may cause intolerant inference latency and GPU memory usage. Existing methods propose mixed-precision quantization to the key-value (KV) cache in LLMs based on token granularity, which is time-consuming in the search process and hardware inefficient during computation. This paper introduces a novel approach called Cocktail, which employs chunk-adaptive mixed-precision quantization to optimize the KV cache. Cocktail consists of two modules: chunk-level quantization sea"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.23294","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/2503.23294/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":"2503.23294","created_at":"2026-07-05T10:41:33.899283+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.23294v1","created_at":"2026-07-05T10:41:33.899283+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.23294","created_at":"2026-07-05T10:41:33.899283+00:00"},{"alias_kind":"pith_short_12","alias_value":"3OU3AZ7GHOPP","created_at":"2026-07-05T10:41:33.899283+00:00"},{"alias_kind":"pith_short_16","alias_value":"3OU3AZ7GHOPP6DMI","created_at":"2026-07-05T10:41:33.899283+00:00"},{"alias_kind":"pith_short_8","alias_value":"3OU3AZ7G","created_at":"2026-07-05T10:41:33.899283+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.20868","citing_title":"Runtime-Certified Bounded-Error Quantized Attention","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6","json":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6.json","graph_json":"https://pith.science/api/pith-number/3OU3AZ7GHOPP6DMIAR5CKXO3K6/graph.json","events_json":"https://pith.science/api/pith-number/3OU3AZ7GHOPP6DMIAR5CKXO3K6/events.json","paper":"https://pith.science/paper/3OU3AZ7G"},"agent_actions":{"view_html":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6","download_json":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6.json","view_paper":"https://pith.science/paper/3OU3AZ7G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.23294&json=true","fetch_graph":"https://pith.science/api/pith-number/3OU3AZ7GHOPP6DMIAR5CKXO3K6/graph.json","fetch_events":"https://pith.science/api/pith-number/3OU3AZ7GHOPP6DMIAR5CKXO3K6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6/action/storage_attestation","attest_author":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6/action/author_attestation","sign_citation":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6/action/citation_signature","submit_replication":"https://pith.science/pith/3OU3AZ7GHOPP6DMIAR5CKXO3K6/action/replication_record"}},"created_at":"2026-07-05T10:41:33.899283+00:00","updated_at":"2026-07-05T10:41:33.899283+00:00"}