{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NLEIIDRUBRMXYGCMN5SYHYFJOH","short_pith_number":"pith:NLEIIDRU","schema_version":"1.0","canonical_sha256":"6ac8840e340c597c184c6f6583e0a971d579f4f2326383faaa8f73926e450b59","source":{"kind":"arxiv","id":"2408.04093","version":4},"attestation_state":"computed","paper":{"title":"Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Beren Millidge, Emily Shepperd, Jonathan Pilault, Quentin Anthony, Vasudev Shyam","submitted_at":"2024-08-07T21:16:55Z","abstract_excerpt":"Our formulation reveals that the reduction across the sequence axis can be efficiently computed in parallel through a tree reduction. Our algorithm, called Tree Attention, for parallelizing exact attention computation across multiple GPUs enables cross-device decoding to be performed asymptotically faster (up to 8x faster in our experiments) than state-of-the-art approaches such as Ring Attention, while also requiring significantly less communication volume and incurring 2x less peak memory. We demonstrate that Tree Attention speeds up decoding up to 4x on Llama 3.1-8B and can be applied to a "},"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":"2408.04093","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-08-07T21:16:55Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"b94fd88618ad95696d2a86a049cd83ce5b96fdbf005f11b8bbcea1fba1484e8c","abstract_canon_sha256":"378e7dc7965ea1208b25203261c5562cc54f28748a5258c2634c8deba3efa050"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:11:22.974835Z","signature_b64":"ly2zrYr1/953PSK7/tJonhC078NEKHfNX72aIvoo+VzXkZjBk2wwa+k2pK08+YSN/+O0piIWllNswbIuys2lCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6ac8840e340c597c184c6f6583e0a971d579f4f2326383faaa8f73926e450b59","last_reissued_at":"2026-07-05T10:11:22.974299Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:11:22.974299Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Beren Millidge, Emily Shepperd, Jonathan Pilault, Quentin Anthony, Vasudev Shyam","submitted_at":"2024-08-07T21:16:55Z","abstract_excerpt":"Our formulation reveals that the reduction across the sequence axis can be efficiently computed in parallel through a tree reduction. Our algorithm, called Tree Attention, for parallelizing exact attention computation across multiple GPUs enables cross-device decoding to be performed asymptotically faster (up to 8x faster in our experiments) than state-of-the-art approaches such as Ring Attention, while also requiring significantly less communication volume and incurring 2x less peak memory. We demonstrate that Tree Attention speeds up decoding up to 4x on Llama 3.1-8B and can be applied to a "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.04093","kind":"arxiv","version":4},"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/2408.04093/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":"2408.04093","created_at":"2026-07-05T10:11:22.974369+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.04093v4","created_at":"2026-07-05T10:11:22.974369+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.04093","created_at":"2026-07-05T10:11:22.974369+00:00"},{"alias_kind":"pith_short_12","alias_value":"NLEIIDRUBRMX","created_at":"2026-07-05T10:11:22.974369+00:00"},{"alias_kind":"pith_short_16","alias_value":"NLEIIDRUBRMXYGCM","created_at":"2026-07-05T10:11:22.974369+00:00"},{"alias_kind":"pith_short_8","alias_value":"NLEIIDRU","created_at":"2026-07-05T10:11:22.974369+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24320","citing_title":"ZONOS2 Technical Report","ref_index":174,"is_internal_anchor":false},{"citing_arxiv_id":"2606.24320","citing_title":"ZONOS2 Technical Report","ref_index":174,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05365","citing_title":"ZAYA1-8B Technical Report","ref_index":149,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH","json":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH.json","graph_json":"https://pith.science/api/pith-number/NLEIIDRUBRMXYGCMN5SYHYFJOH/graph.json","events_json":"https://pith.science/api/pith-number/NLEIIDRUBRMXYGCMN5SYHYFJOH/events.json","paper":"https://pith.science/paper/NLEIIDRU"},"agent_actions":{"view_html":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH","download_json":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH.json","view_paper":"https://pith.science/paper/NLEIIDRU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.04093&json=true","fetch_graph":"https://pith.science/api/pith-number/NLEIIDRUBRMXYGCMN5SYHYFJOH/graph.json","fetch_events":"https://pith.science/api/pith-number/NLEIIDRUBRMXYGCMN5SYHYFJOH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH/action/storage_attestation","attest_author":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH/action/author_attestation","sign_citation":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH/action/citation_signature","submit_replication":"https://pith.science/pith/NLEIIDRUBRMXYGCMN5SYHYFJOH/action/replication_record"}},"created_at":"2026-07-05T10:11:22.974369+00:00","updated_at":"2026-07-05T10:11:22.974369+00:00"}