{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FSAV3SZGWCFL3LVHCVEO4KMZQA","short_pith_number":"pith:FSAV3SZG","schema_version":"1.0","canonical_sha256":"2c815dcb26b08abdaea71548ee299980103b3e0cc73f4f710190dd622dd97691","source":{"kind":"arxiv","id":"2406.11430","version":4},"attestation_state":"computed","paper":{"title":"A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Alessio Devoto, Pasquale Minervini, Simone Scardapane, Yu Zhao","submitted_at":"2024-06-17T11:35:16Z","abstract_excerpt":"The deployment of large language models (LLMs) is often hindered by the extensive memory requirements of the Key-Value (KV) cache, especially as context lengths increase. Existing approaches to reduce the KV cache size involve either fine-tuning the model to learn a compression strategy or leveraging attention scores to reduce the sequence length. We analyse the attention distributions in decoder-only Transformers-based models and observe that attention allocation patterns stay consistent across most layers. Surprisingly, we find a clear correlation between the $L_2$ and the attention scores o"},"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":"2406.11430","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-17T11:35:16Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"795d4874d803c8f70e53092ca8226df61e4c57b9e51784b15e42c2d3f52480f0","abstract_canon_sha256":"6aaca9c79365b2fcdcf11a24d3e9beb1b8178d05422c646264165564b51e943c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:30:17.328226Z","signature_b64":"9Z9Z1pXy65GeSPXCfgCyTWBC+7eaK1wrCL1tgl/jU7ZzkvHlB+FChMnsCBVX8y8uWs3xnaqggu2PPAXw4kgXAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2c815dcb26b08abdaea71548ee299980103b3e0cc73f4f710190dd622dd97691","last_reissued_at":"2026-07-05T09:30:17.327713Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:30:17.327713Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Alessio Devoto, Pasquale Minervini, Simone Scardapane, Yu Zhao","submitted_at":"2024-06-17T11:35:16Z","abstract_excerpt":"The deployment of large language models (LLMs) is often hindered by the extensive memory requirements of the Key-Value (KV) cache, especially as context lengths increase. Existing approaches to reduce the KV cache size involve either fine-tuning the model to learn a compression strategy or leveraging attention scores to reduce the sequence length. We analyse the attention distributions in decoder-only Transformers-based models and observe that attention allocation patterns stay consistent across most layers. Surprisingly, we find a clear correlation between the $L_2$ and the attention scores o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.11430","kind":"arxiv","version":4},"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/2406.11430/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":"2406.11430","created_at":"2026-07-05T09:30:17.327776+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.11430v4","created_at":"2026-07-05T09:30:17.327776+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.11430","created_at":"2026-07-05T09:30:17.327776+00:00"},{"alias_kind":"pith_short_12","alias_value":"FSAV3SZGWCFL","created_at":"2026-07-05T09:30:17.327776+00:00"},{"alias_kind":"pith_short_16","alias_value":"FSAV3SZGWCFL3LVH","created_at":"2026-07-05T09:30:17.327776+00:00"},{"alias_kind":"pith_short_8","alias_value":"FSAV3SZG","created_at":"2026-07-05T09:30:17.327776+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07604","citing_title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","ref_index":154,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07604","citing_title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2410.10781","citing_title":"When Attention Sink Emerges in Language Models: An Empirical View","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25975","citing_title":"Rethinking KV Cache Eviction via a Unified Information-Theoretic Objective","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07234","citing_title":"Reformulating KV Cache Eviction Problem for Long-Context LLM Inference","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA","json":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA.json","graph_json":"https://pith.science/api/pith-number/FSAV3SZGWCFL3LVHCVEO4KMZQA/graph.json","events_json":"https://pith.science/api/pith-number/FSAV3SZGWCFL3LVHCVEO4KMZQA/events.json","paper":"https://pith.science/paper/FSAV3SZG"},"agent_actions":{"view_html":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA","download_json":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA.json","view_paper":"https://pith.science/paper/FSAV3SZG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.11430&json=true","fetch_graph":"https://pith.science/api/pith-number/FSAV3SZGWCFL3LVHCVEO4KMZQA/graph.json","fetch_events":"https://pith.science/api/pith-number/FSAV3SZGWCFL3LVHCVEO4KMZQA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA/action/storage_attestation","attest_author":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA/action/author_attestation","sign_citation":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA/action/citation_signature","submit_replication":"https://pith.science/pith/FSAV3SZGWCFL3LVHCVEO4KMZQA/action/replication_record"}},"created_at":"2026-07-05T09:30:17.327776+00:00","updated_at":"2026-07-05T09:30:17.327776+00:00"}