{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PPZMHLPATI7S74T2BEXNEMJHNQ","short_pith_number":"pith:PPZMHLPA","schema_version":"1.0","canonical_sha256":"7bf2c3ade09a3f2ff27a092ed231276c0ae6ec20a90e252cd898d6d1476f45a1","source":{"kind":"arxiv","id":"2606.22874","version":1},"attestation_state":"computed","paper":{"title":"SpotAttention: Plug-In Block-Sparse Routing for Pretrained Long-Context Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Huzama Ahmad, Se-Young Yun","submitted_at":"2026-06-22T05:39:12Z","abstract_excerpt":"Long contexts have become standard in pretrained LLMs, yet they remain expensive to run: prefill compute grows quadratically with sequence length, and every decode step re-reads a key-value cache that grows linearly with it. Sparse attention cuts these costs by attending only to a relevant subset of past tokens, but selecting that subset is itself expensive. We present SpotAttention, a lightweight selector that attaches to a frozen pretrained transformer and learns by KL distillation to estimate its attention distribution. The selector picks the top-K keys each query attends to, and because it"},"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":"2606.22874","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-22T05:39:12Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0e02c3cdcdd1d79e09c32c2720ac3a6f999d179e16f93dfd8490d8e6f7441751","abstract_canon_sha256":"738fd805d063d472414ac9930c69d8df33e27a22a8c7d5fb1227ba518487fba7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T03:14:02.764978Z","signature_b64":"QggLrF2MrodE4z2GwwKLRHKNa5PARiwFFEX9naC7lW6n9SKXtR4ROrL3KStIAj0h4AJZlPgZbuBZOrjinjkUDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7bf2c3ade09a3f2ff27a092ed231276c0ae6ec20a90e252cd898d6d1476f45a1","last_reissued_at":"2026-06-23T03:14:02.764592Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T03:14:02.764592Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SpotAttention: Plug-In Block-Sparse Routing for Pretrained Long-Context Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Huzama Ahmad, Se-Young Yun","submitted_at":"2026-06-22T05:39:12Z","abstract_excerpt":"Long contexts have become standard in pretrained LLMs, yet they remain expensive to run: prefill compute grows quadratically with sequence length, and every decode step re-reads a key-value cache that grows linearly with it. Sparse attention cuts these costs by attending only to a relevant subset of past tokens, but selecting that subset is itself expensive. We present SpotAttention, a lightweight selector that attaches to a frozen pretrained transformer and learns by KL distillation to estimate its attention distribution. The selector picks the top-K keys each query attends to, and because it"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.22874","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/2606.22874/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":"2606.22874","created_at":"2026-06-23T03:14:02.764649+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.22874v1","created_at":"2026-06-23T03:14:02.764649+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.22874","created_at":"2026-06-23T03:14:02.764649+00:00"},{"alias_kind":"pith_short_12","alias_value":"PPZMHLPATI7S","created_at":"2026-06-23T03:14:02.764649+00:00"},{"alias_kind":"pith_short_16","alias_value":"PPZMHLPATI7S74T2","created_at":"2026-06-23T03:14:02.764649+00:00"},{"alias_kind":"pith_short_8","alias_value":"PPZMHLPA","created_at":"2026-06-23T03:14:02.764649+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/PPZMHLPATI7S74T2BEXNEMJHNQ","json":"https://pith.science/pith/PPZMHLPATI7S74T2BEXNEMJHNQ.json","graph_json":"https://pith.science/api/pith-number/PPZMHLPATI7S74T2BEXNEMJHNQ/graph.json","events_json":"https://pith.science/api/pith-number/PPZMHLPATI7S74T2BEXNEMJHNQ/events.json","paper":"https://pith.science/paper/PPZMHLPA"},"agent_actions":{"view_html":"https://pith.science/pith/PPZMHLPATI7S74T2BEXNEMJHNQ","download_json":"https://pith.science/pith/PPZMHLPATI7S74T2BEXNEMJHNQ.json","view_paper":"https://pith.science/paper/PPZMHLPA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.22874&json=true","fetch_graph":"https://pith.science/api/pith-number/PPZMHLPATI7S74T2BEXNEMJHNQ/graph.json","fetch_events":"https://pith.science/api/pith-number/PPZMHLPATI7S74T2BEXNEMJHNQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PPZMHLPATI7S74T2BEXNEMJHNQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PPZMHLPATI7S74T2BEXNEMJHNQ/action/storage_attestation","attest_author":"https://pith.science/pith/PPZMHLPATI7S74T2BEXNEMJHNQ/action/author_attestation","sign_citation":"https://pith.science/pith/PPZMHLPATI7S74T2BEXNEMJHNQ/action/citation_signature","submit_replication":"https://pith.science/pith/PPZMHLPATI7S74T2BEXNEMJHNQ/action/replication_record"}},"created_at":"2026-06-23T03:14:02.764649+00:00","updated_at":"2026-06-23T03:14:02.764649+00:00"}