{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WHD7RYUSQ2IOUSTMT65V2GGASK","short_pith_number":"pith:WHD7RYUS","schema_version":"1.0","canonical_sha256":"b1c7f8e2928690ea4a6c9fbb5d18c092aba1b448f13f255c2867070c77c3f21d","source":{"kind":"arxiv","id":"2608.08684","version":1},"attestation_state":"computed","paper":{"title":"RippleKV: Cross-Layer KV Cache Allocation via Perturbation Propagation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Dongjie Xu, Fenglei Jin, Hanchi Dong, Jiajie Xu, Julius, Kai Qian, Minghua Tang, Weijie Shi, Yuxuan Sun","submitted_at":"2026-08-09T13:01:11Z","abstract_excerpt":"Long-context LLM inference is bottlenecked by KV cache memory, yet distributing a limited cache budget across layers remains challenging. Existing methods rely on proxies such as layer depth, attention statistics, or representation change. These proxies do not measure how perturbations at each layer propagate to the output and may therefore cause sensitive layers to be underallocated while tolerant layers are overallocated. To address this issue, we propose RippleKV, which allocates cache across layers by estimating how perturbations to each layer's value cache affect the final predictive dist"},"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":"2608.08684","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-09T13:01:11Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"79252a75c4a1ca9317a9711d0c81ce8e6df4c0f8e0d59370078925b36b50cf66","abstract_canon_sha256":"3d9446be9733c181d2fb6c1b2a46468072fdbda108f372a195ac767e0b853621"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-11T01:24:59.969689Z","signature_b64":"gv9JOv7TnZMaNOlppEl6DuxeCNB+yjljI0iZYCF/MD4AQmyI0wik2teuTDbjUbqtpPJqmfCtneQue2ZmmXMODg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b1c7f8e2928690ea4a6c9fbb5d18c092aba1b448f13f255c2867070c77c3f21d","last_reissued_at":"2026-08-11T01:24:59.967226Z","signature_status":"signed_v1","first_computed_at":"2026-08-11T01:24:59.967226Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RippleKV: Cross-Layer KV Cache Allocation via Perturbation Propagation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Dongjie Xu, Fenglei Jin, Hanchi Dong, Jiajie Xu, Julius, Kai Qian, Minghua Tang, Weijie Shi, Yuxuan Sun","submitted_at":"2026-08-09T13:01:11Z","abstract_excerpt":"Long-context LLM inference is bottlenecked by KV cache memory, yet distributing a limited cache budget across layers remains challenging. Existing methods rely on proxies such as layer depth, attention statistics, or representation change. These proxies do not measure how perturbations at each layer propagate to the output and may therefore cause sensitive layers to be underallocated while tolerant layers are overallocated. To address this issue, we propose RippleKV, which allocates cache across layers by estimating how perturbations to each layer's value cache affect the final predictive dist"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.08684","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/2608.08684/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":"2608.08684","created_at":"2026-08-11T01:24:59.968107+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.08684v1","created_at":"2026-08-11T01:24:59.968107+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.08684","created_at":"2026-08-11T01:24:59.968107+00:00"},{"alias_kind":"pith_short_12","alias_value":"WHD7RYUSQ2IO","created_at":"2026-08-11T01:24:59.968107+00:00"},{"alias_kind":"pith_short_16","alias_value":"WHD7RYUSQ2IOUSTM","created_at":"2026-08-11T01:24:59.968107+00:00"},{"alias_kind":"pith_short_8","alias_value":"WHD7RYUS","created_at":"2026-08-11T01:24:59.968107+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK","json":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK.json","graph_json":"https://pith.science/api/pith-number/WHD7RYUSQ2IOUSTMT65V2GGASK/graph.json","events_json":"https://pith.science/api/pith-number/WHD7RYUSQ2IOUSTMT65V2GGASK/events.json","paper":"https://pith.science/paper/WHD7RYUS"},"agent_actions":{"view_html":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK","download_json":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK.json","view_paper":"https://pith.science/paper/WHD7RYUS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.08684&json=true","fetch_graph":"https://pith.science/api/pith-number/WHD7RYUSQ2IOUSTMT65V2GGASK/graph.json","fetch_events":"https://pith.science/api/pith-number/WHD7RYUSQ2IOUSTMT65V2GGASK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK/action/storage_attestation","attest_author":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK/action/author_attestation","sign_citation":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK/action/citation_signature","submit_replication":"https://pith.science/pith/WHD7RYUSQ2IOUSTMT65V2GGASK/action/replication_record"}},"created_at":"2026-08-11T01:24:59.968107+00:00","updated_at":"2026-08-11T01:24:59.968107+00:00"}