{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:3N5BVIFTIHAYG3D6KNJQYDO7GU","short_pith_number":"pith:3N5BVIFT","schema_version":"1.0","canonical_sha256":"db7a1aa0b341c1836c7e53530c0ddf351325d374f1126575d717dd58af44663c","source":{"kind":"arxiv","id":"2407.12820","version":2},"attestation_state":"computed","paper":{"title":"PQCache: Product Quantization-based KVCache for Long Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bin Cui, Fangcheng Fu, Hailin Zhang, Weipeng Chen, Xiaodong Ji, Xiaonan Nie, Xupeng Miao, Yilin Chen","submitted_at":"2024-07-01T13:05:42Z","abstract_excerpt":"As the field of Large Language Models (LLMs) continues to evolve, the context length in inference is steadily growing. Key-Value Cache (KVCache), the intermediate representations of tokens within LLM inference, has now become the primary memory bottleneck due to limited GPU memory. Current methods selectively determine suitable keys and values for self-attention computation in LLMs to address the issue. However, they either fall short in maintaining model quality or result in high serving latency. Drawing inspiration from advanced embedding retrieval techniques prevalent in the data management"},"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":"2407.12820","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-01T13:05:42Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"31a8e66fb1ffcdd10a14300c1eb9e87778933b38adaa39bebd8adb76f298171b","abstract_canon_sha256":"8edd66fc9e5d651db919182ebcd83c4b7c70814c2ca56d83f3814fe5120cc203"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:41:13.861888Z","signature_b64":"hssGUFDZ8CAHfvGu0SERlhtJhvRxzqerD5rfs2X10LX2zE8DoD83+vwCr62LuG4yPsvXRyE+vKYAo4v1IuxSAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"db7a1aa0b341c1836c7e53530c0ddf351325d374f1126575d717dd58af44663c","last_reissued_at":"2026-07-05T10:41:13.861424Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:41:13.861424Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PQCache: Product Quantization-based KVCache for Long Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bin Cui, Fangcheng Fu, Hailin Zhang, Weipeng Chen, Xiaodong Ji, Xiaonan Nie, Xupeng Miao, Yilin Chen","submitted_at":"2024-07-01T13:05:42Z","abstract_excerpt":"As the field of Large Language Models (LLMs) continues to evolve, the context length in inference is steadily growing. Key-Value Cache (KVCache), the intermediate representations of tokens within LLM inference, has now become the primary memory bottleneck due to limited GPU memory. Current methods selectively determine suitable keys and values for self-attention computation in LLMs to address the issue. However, they either fall short in maintaining model quality or result in high serving latency. Drawing inspiration from advanced embedding retrieval techniques prevalent in the data management"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.12820","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/2407.12820/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":"2407.12820","created_at":"2026-07-05T10:41:13.861480+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.12820v2","created_at":"2026-07-05T10:41:13.861480+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.12820","created_at":"2026-07-05T10:41:13.861480+00:00"},{"alias_kind":"pith_short_12","alias_value":"3N5BVIFTIHAY","created_at":"2026-07-05T10:41:13.861480+00:00"},{"alias_kind":"pith_short_16","alias_value":"3N5BVIFTIHAYG3D6","created_at":"2026-07-05T10:41:13.861480+00:00"},{"alias_kind":"pith_short_8","alias_value":"3N5BVIFT","created_at":"2026-07-05T10:41:13.861480+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2409.10516","citing_title":"RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval","ref_index":118,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18137","citing_title":"AQPIM: Breaking the PIM Capacity Wall for LLMs with In-Memory Activation Quantization","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10539","citing_title":"IceCache: Memory-efficient KV-cache Management for Long-Sequence LLMs","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU","json":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU.json","graph_json":"https://pith.science/api/pith-number/3N5BVIFTIHAYG3D6KNJQYDO7GU/graph.json","events_json":"https://pith.science/api/pith-number/3N5BVIFTIHAYG3D6KNJQYDO7GU/events.json","paper":"https://pith.science/paper/3N5BVIFT"},"agent_actions":{"view_html":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU","download_json":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU.json","view_paper":"https://pith.science/paper/3N5BVIFT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.12820&json=true","fetch_graph":"https://pith.science/api/pith-number/3N5BVIFTIHAYG3D6KNJQYDO7GU/graph.json","fetch_events":"https://pith.science/api/pith-number/3N5BVIFTIHAYG3D6KNJQYDO7GU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU/action/storage_attestation","attest_author":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU/action/author_attestation","sign_citation":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU/action/citation_signature","submit_replication":"https://pith.science/pith/3N5BVIFTIHAYG3D6KNJQYDO7GU/action/replication_record"}},"created_at":"2026-07-05T10:41:13.861480+00:00","updated_at":"2026-07-05T10:41:13.861480+00:00"}