{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LSPXIHOAWBYYTAVPBOGNFXDVPE","short_pith_number":"pith:LSPXIHOA","schema_version":"1.0","canonical_sha256":"5c9f741dc0b0718982af0b8cd2dc75792d6a3594b43fcb8e98e874e80ad5d47e","source":{"kind":"arxiv","id":"2402.18096","version":1},"attestation_state":"computed","paper":{"title":"No Token Left Behind: Reliable KV Cache Compression via Importance-Aware Mixed Precision Quantization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Beomseok Kwon, Byeongwook Kim, Dongsoo Lee, Eunho Yang, Gunho Park, Jeongin Bae, June Yong Yang, Se Jung Kwon","submitted_at":"2024-02-28T06:34:54Z","abstract_excerpt":"Key-Value (KV) Caching has become an essential technique for accelerating the inference speed and throughput of generative Large Language Models~(LLMs). However, the memory footprint of the KV cache poses a critical bottleneck in LLM deployment as the cache size grows with batch size and sequence length, often surpassing even the size of the model itself. Although recent methods were proposed to select and evict unimportant KV pairs from the cache to reduce memory consumption, the potential ramifications of eviction on the generative process are yet to be thoroughly examined. In this paper, we"},"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":"2402.18096","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-28T06:34:54Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"010bbfadad32b71b0396bfdba5695a164c3265c1ee4863f33e4a438500a796d4","abstract_canon_sha256":"d78e6b7c0fc7f75315e6de9ed0c65ac6e8cf0fc435aeda1a245caed04fd820d7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:50:12.292162Z","signature_b64":"TscoTGBt8wCvRo6Kk+3X3wFkfIvQxaGJMc7xwgIW5JhtRpnf7zOrnZSakVAnvsRRt7Bpf5pZM31EGV46G9/hAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5c9f741dc0b0718982af0b8cd2dc75792d6a3594b43fcb8e98e874e80ad5d47e","last_reissued_at":"2026-07-05T07:50:12.291562Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:50:12.291562Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"No Token Left Behind: Reliable KV Cache Compression via Importance-Aware Mixed Precision Quantization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Beomseok Kwon, Byeongwook Kim, Dongsoo Lee, Eunho Yang, Gunho Park, Jeongin Bae, June Yong Yang, Se Jung Kwon","submitted_at":"2024-02-28T06:34:54Z","abstract_excerpt":"Key-Value (KV) Caching has become an essential technique for accelerating the inference speed and throughput of generative Large Language Models~(LLMs). However, the memory footprint of the KV cache poses a critical bottleneck in LLM deployment as the cache size grows with batch size and sequence length, often surpassing even the size of the model itself. Although recent methods were proposed to select and evict unimportant KV pairs from the cache to reduce memory consumption, the potential ramifications of eviction on the generative process are yet to be thoroughly examined. In this paper, we"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.18096","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/2402.18096/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":"2402.18096","created_at":"2026-07-05T07:50:12.291678+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.18096v1","created_at":"2026-07-05T07:50:12.291678+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.18096","created_at":"2026-07-05T07:50:12.291678+00:00"},{"alias_kind":"pith_short_12","alias_value":"LSPXIHOAWBYY","created_at":"2026-07-05T07:50:12.291678+00:00"},{"alias_kind":"pith_short_16","alias_value":"LSPXIHOAWBYYTAVP","created_at":"2026-07-05T07:50:12.291678+00:00"},{"alias_kind":"pith_short_8","alias_value":"LSPXIHOA","created_at":"2026-07-05T07:50:12.291678+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24033","citing_title":"RoPE-Aware Bit Allocation for KV-Cache Quantization","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05698","citing_title":"Rethinking LoRA Memory Through the Lens of KV Cache Compression","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31519","citing_title":"RaBitQCache: Rotated Binary Quantization for KVCache in Long Context LLM Inference","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22337","citing_title":"Meta-Soft: Leveraging Composable Meta-Tokens for Context-Preserving KV Cache Compression","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2502.01068","citing_title":"FastKV: Decoupling of Context Reduction and KV Cache Compression for Prefill-Decoding Acceleration","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22337","citing_title":"Meta-Soft: Leveraging Composable Meta-Tokens for Context-Preserving KV Cache Compression","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2504.19874","citing_title":"TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2507.21433","citing_title":"ReasonCache: Accelerating Large Reasoning Model Serving through KV Cache Sharing","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12464","citing_title":"Search Your Block Floating Point Scales!","ref_index":164,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08317","citing_title":"RDKV: Rate-Distortion Bit Allocation for Joint Eviction and Quantization of the KV Cache","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08234","citing_title":"When Does Value-Aware KV Eviction Help? A Fixed-Contract Diagnostic for Non-Monotone Cache Compression","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02262","citing_title":"WindowQuant: Mixed-Precision KV Cache Quantization based on Window-Level Similarity for VLMs Inference Optimization","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE","json":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE.json","graph_json":"https://pith.science/api/pith-number/LSPXIHOAWBYYTAVPBOGNFXDVPE/graph.json","events_json":"https://pith.science/api/pith-number/LSPXIHOAWBYYTAVPBOGNFXDVPE/events.json","paper":"https://pith.science/paper/LSPXIHOA"},"agent_actions":{"view_html":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE","download_json":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE.json","view_paper":"https://pith.science/paper/LSPXIHOA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.18096&json=true","fetch_graph":"https://pith.science/api/pith-number/LSPXIHOAWBYYTAVPBOGNFXDVPE/graph.json","fetch_events":"https://pith.science/api/pith-number/LSPXIHOAWBYYTAVPBOGNFXDVPE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE/action/storage_attestation","attest_author":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE/action/author_attestation","sign_citation":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE/action/citation_signature","submit_replication":"https://pith.science/pith/LSPXIHOAWBYYTAVPBOGNFXDVPE/action/replication_record"}},"created_at":"2026-07-05T07:50:12.291678+00:00","updated_at":"2026-07-05T07:50:12.291678+00:00"}