{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BNSYW7MRXBNSLOFDODW2DB4WZY","short_pith_number":"pith:BNSYW7MR","schema_version":"1.0","canonical_sha256":"0b658b7d91b85b25b8a370eda18796ce1e1c8360163807ca636e8f73833b0edd","source":{"kind":"arxiv","id":"2406.12335","version":2},"attestation_state":"computed","paper":{"title":"Attention Score is not All You Need for Token Importance Indicator in KV Cache Reduction: Value Also Matters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Hidetaka Kamigaito, Taro Watanabe, Zhiyu Guo","submitted_at":"2024-06-18T07:01:11Z","abstract_excerpt":"Scaling the context size of large language models (LLMs) enables them to perform various new tasks, e.g., book summarization. However, the memory cost of the Key and Value (KV) cache in attention significantly limits the practical applications of LLMs. Recent works have explored token pruning for KV cache reduction in LLMs, relying solely on attention scores as a token importance indicator. However, our investigation into value vector norms revealed a notably non-uniform pattern questioning their reliance only on attention scores. Inspired by this, we propose a new method: Value-Aware Token Pr"},"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":"2406.12335","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-18T07:01:11Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"fa48fdc71ae44a43f68e0501fa7ffa7f36099d0e4bcf3ffe7f14de13495769cd","abstract_canon_sha256":"33f54fb1ef3338ef280c6e12ea85bab39deba2ff3a84c986d5806b452b6aef01"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:14:34.684547Z","signature_b64":"IcYdqDg8cWQRPB2uuQU9vvwsexPwiNVcwnFkC0aNg89zLeocz0nL49g8wMTJSLBOpJ4FyftzF8jhTpWNl+q2Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0b658b7d91b85b25b8a370eda18796ce1e1c8360163807ca636e8f73833b0edd","last_reissued_at":"2026-07-05T09:14:34.684013Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:14:34.684013Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Attention Score is not All You Need for Token Importance Indicator in KV Cache Reduction: Value Also Matters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Hidetaka Kamigaito, Taro Watanabe, Zhiyu Guo","submitted_at":"2024-06-18T07:01:11Z","abstract_excerpt":"Scaling the context size of large language models (LLMs) enables them to perform various new tasks, e.g., book summarization. However, the memory cost of the Key and Value (KV) cache in attention significantly limits the practical applications of LLMs. Recent works have explored token pruning for KV cache reduction in LLMs, relying solely on attention scores as a token importance indicator. However, our investigation into value vector norms revealed a notably non-uniform pattern questioning their reliance only on attention scores. Inspired by this, we propose a new method: Value-Aware Token Pr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.12335","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/2406.12335/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":"2406.12335","created_at":"2026-07-05T09:14:34.684072+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.12335v2","created_at":"2026-07-05T09:14:34.684072+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.12335","created_at":"2026-07-05T09:14:34.684072+00:00"},{"alias_kind":"pith_short_12","alias_value":"BNSYW7MRXBNS","created_at":"2026-07-05T09:14:34.684072+00:00"},{"alias_kind":"pith_short_16","alias_value":"BNSYW7MRXBNSLOFD","created_at":"2026-07-05T09:14:34.684072+00:00"},{"alias_kind":"pith_short_8","alias_value":"BNSYW7MR","created_at":"2026-07-05T09:14:34.684072+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.18103","citing_title":"Stability Implies Redundancy: Delta Attention Selective Halting for Efficient Long-Context Prefilling","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2606.07604","citing_title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","ref_index":161,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07604","citing_title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2502.01941","citing_title":"Semantic Integrity Matters: Benchmarking and Preserving High-Density Reasoning in KV Cache Compression","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2602.08686","citing_title":"CompilerKV: Risk-Adaptive KV Compression via Offline Experience Compilation","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2410.10781","citing_title":"When Attention Sink Emerges in Language Models: An Empirical View","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08840","citing_title":"ReST-KV: Robust KV Cache Eviction with Layer-wise Output Reconstruction and Spatial-Temporal Smoothing","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY","json":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY.json","graph_json":"https://pith.science/api/pith-number/BNSYW7MRXBNSLOFDODW2DB4WZY/graph.json","events_json":"https://pith.science/api/pith-number/BNSYW7MRXBNSLOFDODW2DB4WZY/events.json","paper":"https://pith.science/paper/BNSYW7MR"},"agent_actions":{"view_html":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY","download_json":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY.json","view_paper":"https://pith.science/paper/BNSYW7MR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.12335&json=true","fetch_graph":"https://pith.science/api/pith-number/BNSYW7MRXBNSLOFDODW2DB4WZY/graph.json","fetch_events":"https://pith.science/api/pith-number/BNSYW7MRXBNSLOFDODW2DB4WZY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY/action/storage_attestation","attest_author":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY/action/author_attestation","sign_citation":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY/action/citation_signature","submit_replication":"https://pith.science/pith/BNSYW7MRXBNSLOFDODW2DB4WZY/action/replication_record"}},"created_at":"2026-07-05T09:14:34.684072+00:00","updated_at":"2026-07-05T09:14:34.684072+00:00"}