{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SGYRAED7FKYV5XYGV3V7POFIQW","short_pith_number":"pith:SGYRAED7","schema_version":"1.0","canonical_sha256":"91b110107f2ab15edf06aeebf7b8a885aeb00904dc5770136aeff69a39b4ade5","source":{"kind":"arxiv","id":"2510.07651","version":2},"attestation_state":"computed","paper":{"title":"OBCache: Optimal Brain KV Cache Pruning for Efficient Long-Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Enmao Diao, Jiaojiao Zhao, Xiyu Liang, Yuzhe Gu","submitted_at":"2025-10-09T00:58:28Z","abstract_excerpt":"Large language models (LLMs) with extended context windows enable powerful applications but impose significant memory overhead, as caching all key-value (KV) states scales linearly with sequence length and batch size. Existing cache eviction methods address this by exploiting attention sparsity, yet they typically rank tokens heuristically using accumulated attention weights without considering their true impact on attention outputs. We propose Optimal Brain Cache (OBCache), a principled framework that formulates cache eviction as a layer-wise structured pruning problem. Building upon the Opti"},"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":"2510.07651","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-10-09T00:58:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"4e4e09e9abfc244653ceb55493d0aa692f0d8bbf0fe70bb945a0991a3a8c0575","abstract_canon_sha256":"683dabf08ccb66a07c0b8930e50bbdcabed54f656273e22c159bbd22a6f15f31"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:02:23.250223Z","signature_b64":"a+zWsCIy2bhnXFPm1D5sdpONxhSU7NG10DYWdJGvgKVGaSXK4lmopHiG8ftByRKDGmFAXdJypoqapx2MN926CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"91b110107f2ab15edf06aeebf7b8a885aeb00904dc5770136aeff69a39b4ade5","last_reissued_at":"2026-06-01T01:02:23.249020Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:02:23.249020Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"OBCache: Optimal Brain KV Cache Pruning for Efficient Long-Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Enmao Diao, Jiaojiao Zhao, Xiyu Liang, Yuzhe Gu","submitted_at":"2025-10-09T00:58:28Z","abstract_excerpt":"Large language models (LLMs) with extended context windows enable powerful applications but impose significant memory overhead, as caching all key-value (KV) states scales linearly with sequence length and batch size. Existing cache eviction methods address this by exploiting attention sparsity, yet they typically rank tokens heuristically using accumulated attention weights without considering their true impact on attention outputs. We propose Optimal Brain Cache (OBCache), a principled framework that formulates cache eviction as a layer-wise structured pruning problem. Building upon the Opti"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.07651","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/2510.07651/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":"2510.07651","created_at":"2026-06-01T01:02:23.249144+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.07651v2","created_at":"2026-06-01T01:02:23.249144+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.07651","created_at":"2026-06-01T01:02:23.249144+00:00"},{"alias_kind":"pith_short_12","alias_value":"SGYRAED7FKYV","created_at":"2026-06-01T01:02:23.249144+00:00"},{"alias_kind":"pith_short_16","alias_value":"SGYRAED7FKYV5XYG","created_at":"2026-06-01T01:02:23.249144+00:00"},{"alias_kind":"pith_short_8","alias_value":"SGYRAED7","created_at":"2026-06-01T01:02:23.249144+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.08317","citing_title":"RDKV: Rate-Distortion Bit Allocation for Joint Eviction and Quantization of the KV Cache","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2604.10098","citing_title":"Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation","ref_index":37,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW","json":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW.json","graph_json":"https://pith.science/api/pith-number/SGYRAED7FKYV5XYGV3V7POFIQW/graph.json","events_json":"https://pith.science/api/pith-number/SGYRAED7FKYV5XYGV3V7POFIQW/events.json","paper":"https://pith.science/paper/SGYRAED7"},"agent_actions":{"view_html":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW","download_json":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW.json","view_paper":"https://pith.science/paper/SGYRAED7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.07651&json=true","fetch_graph":"https://pith.science/api/pith-number/SGYRAED7FKYV5XYGV3V7POFIQW/graph.json","fetch_events":"https://pith.science/api/pith-number/SGYRAED7FKYV5XYGV3V7POFIQW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW/action/storage_attestation","attest_author":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW/action/author_attestation","sign_citation":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW/action/citation_signature","submit_replication":"https://pith.science/pith/SGYRAED7FKYV5XYGV3V7POFIQW/action/replication_record"}},"created_at":"2026-06-01T01:02:23.249144+00:00","updated_at":"2026-06-01T01:02:23.249144+00:00"}