{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:GG74ICDYQ6MLBM3TTSUW6UKIWD","short_pith_number":"pith:GG74ICDY","schema_version":"1.0","canonical_sha256":"31bfc408788798b0b3739ca96f5148b0e3ca947dc2924b1ea84420a756eb7563","source":{"kind":"arxiv","id":"2605.27740","version":1},"attestation_state":"computed","paper":{"title":"UNIQUE: Universal Top-k Sparse Attention for Training-free Inference and Sparsity-aware Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jinyu Li, Keqi Deng, Ruchao Fan, Shaoshi Ling","submitted_at":"2026-05-26T22:32:43Z","abstract_excerpt":"Long-context inference in large language models (LLMs) is bottlenecked by the linear growth of the self-attention key-value (KV) cache. Top-k sparse attention alleviates this by loading only a small fraction of the KV cache, but accurately and cheaply estimating cache importance, for both training-free use and sparsity-aware training, remains challenging. This paper proposes UNIQUE, a universal top-k sparse attention framework that addresses both requirements and stays consistently effective across LLM modalities. UNIQUE operates at the granularity of KV pages and estimates per-page importance"},"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":"2605.27740","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-26T22:32:43Z","cross_cats_sorted":[],"title_canon_sha256":"49b4c5260348afa0fd8f3599d97dfd8b3735f831e409d9a6eea7bd8e4f45074d","abstract_canon_sha256":"de340dfcb64fd899bd2f536dc66618a1d8806fca0607e50f47fb051321396573"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:47.656296Z","signature_b64":"EFO+jdaJRFIdlhPY9B7op60AvOoMXG7pfu0Jgl36uBrXDfANTa6RbEGa6LsnpYXS7WEZZrmXiGBiTjyqGrA8CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"31bfc408788798b0b3739ca96f5148b0e3ca947dc2924b1ea84420a756eb7563","last_reissued_at":"2026-05-28T01:04:47.655955Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:47.655955Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"UNIQUE: Universal Top-k Sparse Attention for Training-free Inference and Sparsity-aware Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jinyu Li, Keqi Deng, Ruchao Fan, Shaoshi Ling","submitted_at":"2026-05-26T22:32:43Z","abstract_excerpt":"Long-context inference in large language models (LLMs) is bottlenecked by the linear growth of the self-attention key-value (KV) cache. Top-k sparse attention alleviates this by loading only a small fraction of the KV cache, but accurately and cheaply estimating cache importance, for both training-free use and sparsity-aware training, remains challenging. This paper proposes UNIQUE, a universal top-k sparse attention framework that addresses both requirements and stays consistently effective across LLM modalities. UNIQUE operates at the granularity of KV pages and estimates per-page importance"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.27740","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/2605.27740/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":"2605.27740","created_at":"2026-05-28T01:04:47.656004+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.27740v1","created_at":"2026-05-28T01:04:47.656004+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.27740","created_at":"2026-05-28T01:04:47.656004+00:00"},{"alias_kind":"pith_short_12","alias_value":"GG74ICDYQ6ML","created_at":"2026-05-28T01:04:47.656004+00:00"},{"alias_kind":"pith_short_16","alias_value":"GG74ICDYQ6MLBM3T","created_at":"2026-05-28T01:04:47.656004+00:00"},{"alias_kind":"pith_short_8","alias_value":"GG74ICDY","created_at":"2026-05-28T01:04:47.656004+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/GG74ICDYQ6MLBM3TTSUW6UKIWD","json":"https://pith.science/pith/GG74ICDYQ6MLBM3TTSUW6UKIWD.json","graph_json":"https://pith.science/api/pith-number/GG74ICDYQ6MLBM3TTSUW6UKIWD/graph.json","events_json":"https://pith.science/api/pith-number/GG74ICDYQ6MLBM3TTSUW6UKIWD/events.json","paper":"https://pith.science/paper/GG74ICDY"},"agent_actions":{"view_html":"https://pith.science/pith/GG74ICDYQ6MLBM3TTSUW6UKIWD","download_json":"https://pith.science/pith/GG74ICDYQ6MLBM3TTSUW6UKIWD.json","view_paper":"https://pith.science/paper/GG74ICDY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.27740&json=true","fetch_graph":"https://pith.science/api/pith-number/GG74ICDYQ6MLBM3TTSUW6UKIWD/graph.json","fetch_events":"https://pith.science/api/pith-number/GG74ICDYQ6MLBM3TTSUW6UKIWD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GG74ICDYQ6MLBM3TTSUW6UKIWD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GG74ICDYQ6MLBM3TTSUW6UKIWD/action/storage_attestation","attest_author":"https://pith.science/pith/GG74ICDYQ6MLBM3TTSUW6UKIWD/action/author_attestation","sign_citation":"https://pith.science/pith/GG74ICDYQ6MLBM3TTSUW6UKIWD/action/citation_signature","submit_replication":"https://pith.science/pith/GG74ICDYQ6MLBM3TTSUW6UKIWD/action/replication_record"}},"created_at":"2026-05-28T01:04:47.656004+00:00","updated_at":"2026-05-28T01:04:47.656004+00:00"}