{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5SUGOE43MTTYBBGS5EQXGH73JL","short_pith_number":"pith:5SUGOE43","schema_version":"1.0","canonical_sha256":"eca867139b64e78084d2e921731ffb4af31058bca1f674d3164da7264cfa142d","source":{"kind":"arxiv","id":"2510.24606","version":2},"attestation_state":"computed","paper":{"title":"Long-Context Modeling with Dynamic Hierarchical Sparse Attention for Memory-Constrained LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Faramarz Fekri, Joe Zou, Siheng Xiong, Yae Jee Cho","submitted_at":"2025-10-28T16:34:18Z","abstract_excerpt":"The quadratic cost of attention limits the scalability of long-context LLMs, especially under limited hardware memory budgets. While attention is often sparse, existing static sparse methods cannot adapt to task- or input-dependent variations, and recent dynamic approaches rely on predefined templates or heuristics that may sacrifice generality. We propose Dynamic Hierarchical Sparse Attention (DHSA), a data-driven framework that predicts attention sparsity online while keeping the LLM backbone frozen. DHSA performs hierarchical routing by estimating importance at the chunk level and propagati"},"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":"2510.24606","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-10-28T16:34:18Z","cross_cats_sorted":[],"title_canon_sha256":"644cb592616be24871b262237c3d7ec1ea99ad1f52efdf82645f0531f1f66d23","abstract_canon_sha256":"0bfc70e870811e15f5135cd5992f44920a55ad1d2e02ff70df2e7567798656c8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T01:04:57.468014Z","signature_b64":"Au6sUpbHcCJzWPS5lYBHEjMGR0fZYuuOz5teTuLOuefuyQcjgFN5GboP8nZ0pbACmUNYHwUwpdqYSVP4FBl2Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eca867139b64e78084d2e921731ffb4af31058bca1f674d3164da7264cfa142d","last_reissued_at":"2026-05-29T01:04:57.467473Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T01:04:57.467473Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Long-Context Modeling with Dynamic Hierarchical Sparse Attention for Memory-Constrained LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Faramarz Fekri, Joe Zou, Siheng Xiong, Yae Jee Cho","submitted_at":"2025-10-28T16:34:18Z","abstract_excerpt":"The quadratic cost of attention limits the scalability of long-context LLMs, especially under limited hardware memory budgets. While attention is often sparse, existing static sparse methods cannot adapt to task- or input-dependent variations, and recent dynamic approaches rely on predefined templates or heuristics that may sacrifice generality. We propose Dynamic Hierarchical Sparse Attention (DHSA), a data-driven framework that predicts attention sparsity online while keeping the LLM backbone frozen. DHSA performs hierarchical routing by estimating importance at the chunk level and propagati"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.24606","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/2510.24606/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":"2510.24606","created_at":"2026-05-29T01:04:57.467532+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.24606v2","created_at":"2026-05-29T01:04:57.467532+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.24606","created_at":"2026-05-29T01:04:57.467532+00:00"},{"alias_kind":"pith_short_12","alias_value":"5SUGOE43MTTY","created_at":"2026-05-29T01:04:57.467532+00:00"},{"alias_kind":"pith_short_16","alias_value":"5SUGOE43MTTYBBGS","created_at":"2026-05-29T01:04:57.467532+00:00"},{"alias_kind":"pith_short_8","alias_value":"5SUGOE43","created_at":"2026-05-29T01:04:57.467532+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/5SUGOE43MTTYBBGS5EQXGH73JL","json":"https://pith.science/pith/5SUGOE43MTTYBBGS5EQXGH73JL.json","graph_json":"https://pith.science/api/pith-number/5SUGOE43MTTYBBGS5EQXGH73JL/graph.json","events_json":"https://pith.science/api/pith-number/5SUGOE43MTTYBBGS5EQXGH73JL/events.json","paper":"https://pith.science/paper/5SUGOE43"},"agent_actions":{"view_html":"https://pith.science/pith/5SUGOE43MTTYBBGS5EQXGH73JL","download_json":"https://pith.science/pith/5SUGOE43MTTYBBGS5EQXGH73JL.json","view_paper":"https://pith.science/paper/5SUGOE43","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.24606&json=true","fetch_graph":"https://pith.science/api/pith-number/5SUGOE43MTTYBBGS5EQXGH73JL/graph.json","fetch_events":"https://pith.science/api/pith-number/5SUGOE43MTTYBBGS5EQXGH73JL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5SUGOE43MTTYBBGS5EQXGH73JL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5SUGOE43MTTYBBGS5EQXGH73JL/action/storage_attestation","attest_author":"https://pith.science/pith/5SUGOE43MTTYBBGS5EQXGH73JL/action/author_attestation","sign_citation":"https://pith.science/pith/5SUGOE43MTTYBBGS5EQXGH73JL/action/citation_signature","submit_replication":"https://pith.science/pith/5SUGOE43MTTYBBGS5EQXGH73JL/action/replication_record"}},"created_at":"2026-05-29T01:04:57.467532+00:00","updated_at":"2026-05-29T01:04:57.467532+00:00"}