{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:54PQIH33LC3572WDOTYQQMPVOL","short_pith_number":"pith:54PQIH33","schema_version":"1.0","canonical_sha256":"ef1f041f7b58b7dfeac374f10831f572d3b144979b423da3fb3b6b1ffc59628e","source":{"kind":"arxiv","id":"2411.18077","version":3},"attestation_state":"computed","paper":{"title":"MiniKV: Pushing the Limits of LLM Inference via 2-Bit Layer-Discriminative KV Cache","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Akshat Sharma, Hangliang Ding, Jianping Li, Minjia Zhang, Neel Dani","submitted_at":"2024-11-27T06:10:49Z","abstract_excerpt":"How to efficiently serve LLMs in practice has become exceptionally challenging due to their prohibitive memory and computation requirements. In this study, we investigate optimizing the KV cache, whose memory footprint poses a critical bottleneck in LLM inference, especially when dealing with long context tasks. To tackle the challenge, we introduce MiniKV, a KV cache optimization method that simultaneously preserves long context task accuracy while significantly reducing KV cache size via a novel 2-bit layer-discriminative KV cache. More importantly, we develop specialized CUDA kernels to mak"},"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":"2411.18077","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-11-27T06:10:49Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"751522503f2642e2abb1a57db4fd24f937a7bc11a6e85f6ccb155257ee92180b","abstract_canon_sha256":"1627bd67d9962427bd2fc07fbaae08ba8fd4f502d9284f3ec8fb7e766d90d416"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:18:16.155551Z","signature_b64":"C3dckIGeARkuXzJqgJeqq9vFyIvWIvZ07pGKROtal4DnVqPpVpJNXrBMlWZ1m0jtxa+VWAuSJdhDa34DMmjQDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ef1f041f7b58b7dfeac374f10831f572d3b144979b423da3fb3b6b1ffc59628e","last_reissued_at":"2026-07-05T11:18:16.154991Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:18:16.154991Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MiniKV: Pushing the Limits of LLM Inference via 2-Bit Layer-Discriminative KV Cache","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Akshat Sharma, Hangliang Ding, Jianping Li, Minjia Zhang, Neel Dani","submitted_at":"2024-11-27T06:10:49Z","abstract_excerpt":"How to efficiently serve LLMs in practice has become exceptionally challenging due to their prohibitive memory and computation requirements. In this study, we investigate optimizing the KV cache, whose memory footprint poses a critical bottleneck in LLM inference, especially when dealing with long context tasks. To tackle the challenge, we introduce MiniKV, a KV cache optimization method that simultaneously preserves long context task accuracy while significantly reducing KV cache size via a novel 2-bit layer-discriminative KV cache. More importantly, we develop specialized CUDA kernels to mak"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.18077","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/2411.18077/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":"2411.18077","created_at":"2026-07-05T11:18:16.155059+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.18077v3","created_at":"2026-07-05T11:18:16.155059+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.18077","created_at":"2026-07-05T11:18:16.155059+00:00"},{"alias_kind":"pith_short_12","alias_value":"54PQIH33LC35","created_at":"2026-07-05T11:18:16.155059+00:00"},{"alias_kind":"pith_short_16","alias_value":"54PQIH33LC3572WD","created_at":"2026-07-05T11:18:16.155059+00:00"},{"alias_kind":"pith_short_8","alias_value":"54PQIH33","created_at":"2026-07-05T11:18:16.155059+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.14292","citing_title":"Minimal-Intervention KV Retention via Set-Conditioned Diversity","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17757","citing_title":"OSCAR: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL","json":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL.json","graph_json":"https://pith.science/api/pith-number/54PQIH33LC3572WDOTYQQMPVOL/graph.json","events_json":"https://pith.science/api/pith-number/54PQIH33LC3572WDOTYQQMPVOL/events.json","paper":"https://pith.science/paper/54PQIH33"},"agent_actions":{"view_html":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL","download_json":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL.json","view_paper":"https://pith.science/paper/54PQIH33","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.18077&json=true","fetch_graph":"https://pith.science/api/pith-number/54PQIH33LC3572WDOTYQQMPVOL/graph.json","fetch_events":"https://pith.science/api/pith-number/54PQIH33LC3572WDOTYQQMPVOL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL/action/storage_attestation","attest_author":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL/action/author_attestation","sign_citation":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL/action/citation_signature","submit_replication":"https://pith.science/pith/54PQIH33LC3572WDOTYQQMPVOL/action/replication_record"}},"created_at":"2026-07-05T11:18:16.155059+00:00","updated_at":"2026-07-05T11:18:16.155059+00:00"}