{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:A44N37R3E7JJHYWIIDK6E2DAOK","short_pith_number":"pith:A44N37R3","schema_version":"1.0","canonical_sha256":"0738ddfe3b27d293e2c840d5e26860729778bb30852bf8754304f61c15d11e91","source":{"kind":"arxiv","id":"2412.03131","version":3},"attestation_state":"computed","paper":{"title":"DiffKV: Differentiated Memory Management for Large Language Models with Parallel KV Compaction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Haibo Chen, John C.S. Lui, Runyuan Zhao, Yanqi Zhang, Yuwei Hu","submitted_at":"2024-12-04T08:51:23Z","abstract_excerpt":"Large language models (LLMs) demonstrate remarkable capabilities but face substantial serving costs due to their high memory demands, with the key-value (KV) cache being a primary bottleneck. State-of-the-art KV cache compression techniques, such as quantization and pruning, apply uniform treatment to both keys and values, and discard unimportant tokens entirely, overlooking the fine-grained distinctions in the significance of individual KV cache components. To address such limitations, we introduce \\textit{DiffKV}, a novel framework for efficient KV cache compression that exploits three level"},"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":"2412.03131","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-12-04T08:51:23Z","cross_cats_sorted":["cs.DC"],"title_canon_sha256":"5245cc2088c6ef08499859b00d9b47e0b305d103c9f3478955b6bc4b14bc1071","abstract_canon_sha256":"ff33c44ae3235b69dc21aea4bd7667511604cea3837f20abfef65d09166fde4e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:01:56.165276Z","signature_b64":"n1MswfL/36XO2xW7RCNH2osvFnL893oukr2Scbxv/JNBghis49zHT5SVy/nfwVAa5qfRR1DgABZjNuC3ooWbDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0738ddfe3b27d293e2c840d5e26860729778bb30852bf8754304f61c15d11e91","last_reissued_at":"2026-07-05T12:01:56.164768Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:01:56.164768Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DiffKV: Differentiated Memory Management for Large Language Models with Parallel KV Compaction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Haibo Chen, John C.S. Lui, Runyuan Zhao, Yanqi Zhang, Yuwei Hu","submitted_at":"2024-12-04T08:51:23Z","abstract_excerpt":"Large language models (LLMs) demonstrate remarkable capabilities but face substantial serving costs due to their high memory demands, with the key-value (KV) cache being a primary bottleneck. State-of-the-art KV cache compression techniques, such as quantization and pruning, apply uniform treatment to both keys and values, and discard unimportant tokens entirely, overlooking the fine-grained distinctions in the significance of individual KV cache components. To address such limitations, we introduce \\textit{DiffKV}, a novel framework for efficient KV cache compression that exploits three level"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.03131","kind":"arxiv","version":3},"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/2412.03131/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":"2412.03131","created_at":"2026-07-05T12:01:56.164821+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.03131v3","created_at":"2026-07-05T12:01:56.164821+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.03131","created_at":"2026-07-05T12:01:56.164821+00:00"},{"alias_kind":"pith_short_12","alias_value":"A44N37R3E7JJ","created_at":"2026-07-05T12:01:56.164821+00:00"},{"alias_kind":"pith_short_16","alias_value":"A44N37R3E7JJHYWI","created_at":"2026-07-05T12:01:56.164821+00:00"},{"alias_kind":"pith_short_8","alias_value":"A44N37R3","created_at":"2026-07-05T12:01:56.164821+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08565","citing_title":"SMetric: Rethink LLM Scheduling for Serving Agents with Balanced Session-centric Scheduling","ref_index":57,"is_internal_anchor":true},{"citing_arxiv_id":"2605.03644","citing_title":"AdapShot: Adaptive Many-Shot In-Context Learning with Semantic-Aware KV Cache Reuse","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2505.02922","citing_title":"RetroInfer: A Vector Storage Engine for Scalable Long-Context LLM Inference","ref_index":122,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK","json":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK.json","graph_json":"https://pith.science/api/pith-number/A44N37R3E7JJHYWIIDK6E2DAOK/graph.json","events_json":"https://pith.science/api/pith-number/A44N37R3E7JJHYWIIDK6E2DAOK/events.json","paper":"https://pith.science/paper/A44N37R3"},"agent_actions":{"view_html":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK","download_json":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK.json","view_paper":"https://pith.science/paper/A44N37R3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.03131&json=true","fetch_graph":"https://pith.science/api/pith-number/A44N37R3E7JJHYWIIDK6E2DAOK/graph.json","fetch_events":"https://pith.science/api/pith-number/A44N37R3E7JJHYWIIDK6E2DAOK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK/action/storage_attestation","attest_author":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK/action/author_attestation","sign_citation":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK/action/citation_signature","submit_replication":"https://pith.science/pith/A44N37R3E7JJHYWIIDK6E2DAOK/action/replication_record"}},"created_at":"2026-07-05T12:01:56.164821+00:00","updated_at":"2026-07-05T12:01:56.164821+00:00"}