{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FDDICGXTKSC4XPJ5O45JA65QST","short_pith_number":"pith:FDDICGXT","schema_version":"1.0","canonical_sha256":"28c6811af35485cbbd3d773a907bb094f72d936ce66283da100c8410acb703ad","source":{"kind":"arxiv","id":"2607.06519","version":1},"attestation_state":"computed","paper":{"title":"FreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Adam Puente Tercero, Ainhoa Miranda, Anna C\\'ordoba, Jes\\'us Olivera, Julia Barrientos, Mar Linares Tercero, Nerea Angulo Hijo","submitted_at":"2026-07-07T17:26:28Z","abstract_excerpt":"Long-context LLM inference is increasingly limited by the memory and bandwidth cost of KV caches, yet aggressive compression can remove the layer-specific evidence needed for retrieval and multi-step reasoning. We introduce FreqDepthKV, an inference-time cache compression method that factorizes adjacent-layer KV states into shared low-frequency depth components and sparse high-frequency residuals. A lightweight online probe assigns attention heads to shared-depth, residual-depth, or exact cache modes according to their contribution to reconstruction-sensitive attention logits, allowing the com"},"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":"2607.06519","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-07-07T17:26:28Z","cross_cats_sorted":[],"title_canon_sha256":"90458876b6d320732e9220658258eb80cbcc3ce5254fc521172ebdd507b327d5","abstract_canon_sha256":"bbc0ccb5b626f818b44fd56558b0ec65a4bed1e25f00beb644e12d8861d27616"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-08T01:19:28.675422Z","signature_b64":"kzgggatDB9XpIh6INqpzush1ujGvS6Fm1sVlB/XLixe24cnXgbbNQhFWtztL3jsx9yx4dQuvq4KLbtoKmJqbDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"28c6811af35485cbbd3d773a907bb094f72d936ce66283da100c8410acb703ad","last_reissued_at":"2026-07-08T01:19:28.674983Z","signature_status":"signed_v1","first_computed_at":"2026-07-08T01:19:28.674983Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Adam Puente Tercero, Ainhoa Miranda, Anna C\\'ordoba, Jes\\'us Olivera, Julia Barrientos, Mar Linares Tercero, Nerea Angulo Hijo","submitted_at":"2026-07-07T17:26:28Z","abstract_excerpt":"Long-context LLM inference is increasingly limited by the memory and bandwidth cost of KV caches, yet aggressive compression can remove the layer-specific evidence needed for retrieval and multi-step reasoning. We introduce FreqDepthKV, an inference-time cache compression method that factorizes adjacent-layer KV states into shared low-frequency depth components and sparse high-frequency residuals. A lightweight online probe assigns attention heads to shared-depth, residual-depth, or exact cache modes according to their contribution to reconstruction-sensitive attention logits, allowing the com"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.06519","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/2607.06519/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":"2607.06519","created_at":"2026-07-08T01:19:28.675039+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.06519v1","created_at":"2026-07-08T01:19:28.675039+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.06519","created_at":"2026-07-08T01:19:28.675039+00:00"},{"alias_kind":"pith_short_12","alias_value":"FDDICGXTKSC4","created_at":"2026-07-08T01:19:28.675039+00:00"},{"alias_kind":"pith_short_16","alias_value":"FDDICGXTKSC4XPJ5","created_at":"2026-07-08T01:19:28.675039+00:00"},{"alias_kind":"pith_short_8","alias_value":"FDDICGXT","created_at":"2026-07-08T01:19:28.675039+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/FDDICGXTKSC4XPJ5O45JA65QST","json":"https://pith.science/pith/FDDICGXTKSC4XPJ5O45JA65QST.json","graph_json":"https://pith.science/api/pith-number/FDDICGXTKSC4XPJ5O45JA65QST/graph.json","events_json":"https://pith.science/api/pith-number/FDDICGXTKSC4XPJ5O45JA65QST/events.json","paper":"https://pith.science/paper/FDDICGXT"},"agent_actions":{"view_html":"https://pith.science/pith/FDDICGXTKSC4XPJ5O45JA65QST","download_json":"https://pith.science/pith/FDDICGXTKSC4XPJ5O45JA65QST.json","view_paper":"https://pith.science/paper/FDDICGXT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.06519&json=true","fetch_graph":"https://pith.science/api/pith-number/FDDICGXTKSC4XPJ5O45JA65QST/graph.json","fetch_events":"https://pith.science/api/pith-number/FDDICGXTKSC4XPJ5O45JA65QST/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FDDICGXTKSC4XPJ5O45JA65QST/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FDDICGXTKSC4XPJ5O45JA65QST/action/storage_attestation","attest_author":"https://pith.science/pith/FDDICGXTKSC4XPJ5O45JA65QST/action/author_attestation","sign_citation":"https://pith.science/pith/FDDICGXTKSC4XPJ5O45JA65QST/action/citation_signature","submit_replication":"https://pith.science/pith/FDDICGXTKSC4XPJ5O45JA65QST/action/replication_record"}},"created_at":"2026-07-08T01:19:28.675039+00:00","updated_at":"2026-07-08T01:19:28.675039+00:00"}