{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:UQKLUT5OACZ5Y4RG43GMMQZULG","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"05f2351140b0c95c75fb9dd0b5482630db4c8fc20fffd2e571fba11022538a00","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-07-11T15:44:48Z","title_canon_sha256":"7ccaf2708ba092d27a2f6b45130691fbbd7fbe07983a8bf8e1755befba2fd970"},"schema_version":"1.0","source":{"id":"2407.08608","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2407.08608","created_at":"2026-05-20T19:38:40Z"},{"alias_kind":"arxiv_version","alias_value":"2407.08608v2","created_at":"2026-05-20T19:38:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.08608","created_at":"2026-05-20T19:38:40Z"},{"alias_kind":"pith_short_12","alias_value":"UQKLUT5OACZ5","created_at":"2026-05-20T19:38:40Z"},{"alias_kind":"pith_short_16","alias_value":"UQKLUT5OACZ5Y4RG","created_at":"2026-05-20T19:38:40Z"},{"alias_kind":"pith_short_8","alias_value":"UQKLUT5O","created_at":"2026-05-20T19:38:40Z"}],"graph_snapshots":[{"event_id":"sha256:ac310a1a867327e55a0ab3f7061f13a62355e7056d9ccacb7a955e1e5f27bad1","target":"graph","created_at":"2026-05-20T19:38:40Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2407.08608/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Attention, as a core layer of the ubiquitous Transformer architecture, is the bottleneck for large language models and long-context applications. FlashAttention elaborated an approach to speed up attention on GPUs through minimizing memory reads/writes. However, it has yet to take advantage of new capabilities present in recent hardware, with FlashAttention-2 achieving only 35% utilization on the H100 GPU. We develop three main techniques to speed up attention on Hopper GPUs: exploiting asynchrony of the Tensor Cores and TMA to (1) overlap overall computation and data movement via warp-special","authors_text":"Ganesh Bikshandi, Jay Shah, Pradeep Ramani, Tri Dao, Vijay Thakkar, Ying Zhang","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-07-11T15:44:48Z","title":"FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.08608","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f1c5bca5e9fe7e10cac2ea87678b8f01cb4eff00eec833ee8e1b9f34fdad54d9","target":"record","created_at":"2026-05-20T19:38:40Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"05f2351140b0c95c75fb9dd0b5482630db4c8fc20fffd2e571fba11022538a00","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-07-11T15:44:48Z","title_canon_sha256":"7ccaf2708ba092d27a2f6b45130691fbbd7fbe07983a8bf8e1755befba2fd970"},"schema_version":"1.0","source":{"id":"2407.08608","kind":"arxiv","version":2}},"canonical_sha256":"a414ba4fae00b3dc7226e6ccc64334598e87d37e5ff1092aa3c114c8531c7c68","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a414ba4fae00b3dc7226e6ccc64334598e87d37e5ff1092aa3c114c8531c7c68","first_computed_at":"2026-05-20T19:38:40.310580Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T19:38:40.310580Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"4kFCTH+VbKtdU1jZ/MKcyd8zWXktIkWOTKUSi+8eofNDvKZaC/wf288gJPoY96kzFBDSM/7i3FKJbi25DK3EAA==","signature_status":"signed_v1","signed_at":"2026-05-20T19:38:40.312261Z","signed_message":"canonical_sha256_bytes"},"source_id":"2407.08608","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f1c5bca5e9fe7e10cac2ea87678b8f01cb4eff00eec833ee8e1b9f34fdad54d9","sha256:ac310a1a867327e55a0ab3f7061f13a62355e7056d9ccacb7a955e1e5f27bad1"],"state_sha256":"dc17584f2248f4d771c5aebab0c89f136ce4bbf5e9f54c8e972594e9fdfe6a47"}