{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:3HBLRQPHZ5PT7QT7J6MNCD2PBT","short_pith_number":"pith:3HBLRQPH","canonical_record":{"source":{"id":"2310.03294","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-05T03:47:57Z","cross_cats_sorted":["cs.AI","cs.DC"],"title_canon_sha256":"1e8a35ce44d5773751f8f000baf3afedff78c46e9b6019a392766d6e77929f0b","abstract_canon_sha256":"184a8c5bf20a96648214c8b1ecf0e441da3b0bac29299f046e9f659c5b182393"},"schema_version":"1.0"},"canonical_sha256":"d9c2b8c1e7cf5f3fc27f4f98d10f4f0cef963974c13153f0f3313dbe7442f984","source":{"kind":"arxiv","id":"2310.03294","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2310.03294","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"arxiv_version","alias_value":"2310.03294v2","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.03294","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"pith_short_12","alias_value":"3HBLRQPHZ5PT","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"pith_short_16","alias_value":"3HBLRQPHZ5PT7QT7","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"pith_short_8","alias_value":"3HBLRQPH","created_at":"2026-07-05T08:02:46Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:3HBLRQPHZ5PT7QT7J6MNCD2PBT","target":"record","payload":{"canonical_record":{"source":{"id":"2310.03294","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-05T03:47:57Z","cross_cats_sorted":["cs.AI","cs.DC"],"title_canon_sha256":"1e8a35ce44d5773751f8f000baf3afedff78c46e9b6019a392766d6e77929f0b","abstract_canon_sha256":"184a8c5bf20a96648214c8b1ecf0e441da3b0bac29299f046e9f659c5b182393"},"schema_version":"1.0"},"canonical_sha256":"d9c2b8c1e7cf5f3fc27f4f98d10f4f0cef963974c13153f0f3313dbe7442f984","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:02:46.702707Z","signature_b64":"ezkHmf28zDH3ydHvzQHv3B7m5js9wFCYhg53RCTcrDRtZglzuHJb6k3nf1+wrQ5BsIxJCZakCIcI+X2CfSLlBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d9c2b8c1e7cf5f3fc27f4f98d10f4f0cef963974c13153f0f3313dbe7442f984","last_reissued_at":"2026-07-05T08:02:46.702204Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:02:46.702204Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2310.03294","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:02:46Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"w7W4KTvpgjJRpN9Ha38xOH3TSMBdihiW5VG37cNgcGT68cDSTKKgEWbwqdONb4OKbdQ1b6LrV8dmRsd2Y+qXAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T22:31:03.143697Z"},"content_sha256":"5ff8c99075343a1cc2a93ad29fa6beb5c80969232046679d76e29307ae6e13bf","schema_version":"1.0","event_id":"sha256:5ff8c99075343a1cc2a93ad29fa6beb5c80969232046679d76e29307ae6e13bf"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:3HBLRQPHZ5PT7QT7J6MNCD2PBT","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"DISTFLASHATTN: Distributed Memory-efficient Attention for Long-context LLMs Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.DC"],"primary_cat":"cs.LG","authors_text":"Anze Xie, Dacheng Li, Eric P. Xing, Hao Zhang, Ion Stoica, Joseph E. Gonzalez, Rulin Shao, Xuezhe Ma","submitted_at":"2023-10-05T03:47:57Z","abstract_excerpt":"FlashAttention (Dao, 2023) effectively reduces the quadratic peak memory usage to linear in training transformer-based large language models (LLMs) on a single GPU. In this paper, we introduce DISTFLASHATTN, a distributed memory-efficient attention mechanism optimized for long-context LLMs training. We propose three key techniques: token-level workload balancing, overlapping key-value communication, and a rematerialization-aware gradient checkpointing algorithm. We evaluate DISTFLASHATTN on Llama-7B and variants with sequence lengths from 32K to 512K. DISTFLASHATTN achieves 8x longer sequences"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.03294","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/2310.03294/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:02:46Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"yN8qza4ciJCDCMKMZvI9GBJsOCq3kN9IYeSJjXg7e5NjTVFxj5acDzLkTlPAX8FdU2w+6RVyK4MSZCC+JPT0Cw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T22:31:03.144680Z"},"content_sha256":"04bb8bbb97f58c0e8ee70966c203fe115dbe2b34b0608cba8b179b71f2225d16","schema_version":"1.0","event_id":"sha256:04bb8bbb97f58c0e8ee70966c203fe115dbe2b34b0608cba8b179b71f2225d16"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/3HBLRQPHZ5PT7QT7J6MNCD2PBT/bundle.json","state_url":"https://pith.science/pith/3HBLRQPHZ5PT7QT7J6MNCD2PBT/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/3HBLRQPHZ5PT7QT7J6MNCD2PBT/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-09T22:31:03Z","links":{"resolver":"https://pith.science/pith/3HBLRQPHZ5PT7QT7J6MNCD2PBT","bundle":"https://pith.science/pith/3HBLRQPHZ5PT7QT7J6MNCD2PBT/bundle.json","state":"https://pith.science/pith/3HBLRQPHZ5PT7QT7J6MNCD2PBT/state.json","well_known_bundle":"https://pith.science/.well-known/pith/3HBLRQPHZ5PT7QT7J6MNCD2PBT/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:3HBLRQPHZ5PT7QT7J6MNCD2PBT","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":"184a8c5bf20a96648214c8b1ecf0e441da3b0bac29299f046e9f659c5b182393","cross_cats_sorted":["cs.AI","cs.DC"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-05T03:47:57Z","title_canon_sha256":"1e8a35ce44d5773751f8f000baf3afedff78c46e9b6019a392766d6e77929f0b"},"schema_version":"1.0","source":{"id":"2310.03294","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2310.03294","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"arxiv_version","alias_value":"2310.03294v2","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.03294","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"pith_short_12","alias_value":"3HBLRQPHZ5PT","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"pith_short_16","alias_value":"3HBLRQPHZ5PT7QT7","created_at":"2026-07-05T08:02:46Z"},{"alias_kind":"pith_short_8","alias_value":"3HBLRQPH","created_at":"2026-07-05T08:02:46Z"}],"graph_snapshots":[{"event_id":"sha256:04bb8bbb97f58c0e8ee70966c203fe115dbe2b34b0608cba8b179b71f2225d16","target":"graph","created_at":"2026-07-05T08:02:46Z","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/2310.03294/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"FlashAttention (Dao, 2023) effectively reduces the quadratic peak memory usage to linear in training transformer-based large language models (LLMs) on a single GPU. In this paper, we introduce DISTFLASHATTN, a distributed memory-efficient attention mechanism optimized for long-context LLMs training. We propose three key techniques: token-level workload balancing, overlapping key-value communication, and a rematerialization-aware gradient checkpointing algorithm. We evaluate DISTFLASHATTN on Llama-7B and variants with sequence lengths from 32K to 512K. DISTFLASHATTN achieves 8x longer sequences","authors_text":"Anze Xie, Dacheng Li, Eric P. Xing, Hao Zhang, Ion Stoica, Joseph E. Gonzalez, Rulin Shao, Xuezhe Ma","cross_cats":["cs.AI","cs.DC"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-05T03:47:57Z","title":"DISTFLASHATTN: Distributed Memory-efficient Attention for Long-context LLMs Training"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.03294","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":"sha256:5ff8c99075343a1cc2a93ad29fa6beb5c80969232046679d76e29307ae6e13bf","target":"record","created_at":"2026-07-05T08:02:46Z","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":"184a8c5bf20a96648214c8b1ecf0e441da3b0bac29299f046e9f659c5b182393","cross_cats_sorted":["cs.AI","cs.DC"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-05T03:47:57Z","title_canon_sha256":"1e8a35ce44d5773751f8f000baf3afedff78c46e9b6019a392766d6e77929f0b"},"schema_version":"1.0","source":{"id":"2310.03294","kind":"arxiv","version":2}},"canonical_sha256":"d9c2b8c1e7cf5f3fc27f4f98d10f4f0cef963974c13153f0f3313dbe7442f984","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d9c2b8c1e7cf5f3fc27f4f98d10f4f0cef963974c13153f0f3313dbe7442f984","first_computed_at":"2026-07-05T08:02:46.702204Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:02:46.702204Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ezkHmf28zDH3ydHvzQHv3B7m5js9wFCYhg53RCTcrDRtZglzuHJb6k3nf1+wrQ5BsIxJCZakCIcI+X2CfSLlBA==","signature_status":"signed_v1","signed_at":"2026-07-05T08:02:46.702707Z","signed_message":"canonical_sha256_bytes"},"source_id":"2310.03294","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5ff8c99075343a1cc2a93ad29fa6beb5c80969232046679d76e29307ae6e13bf","sha256:04bb8bbb97f58c0e8ee70966c203fe115dbe2b34b0608cba8b179b71f2225d16"],"state_sha256":"c8e3ca3783083ec042972b20e38aa37e2e33142d103a4dade9cca16d4b6dec85"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tHbMI0luYFfdIZBZ1MYvSRpUi8W28kqLdYdwiNzAZhl0n7ohFdN/DE7z0EETaXqrzRlWwp3JJVScg50OJgaNAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T22:31:03.152193Z","bundle_sha256":"af64d39afa2c26153cf8638c09a2276b9f281e5e3967ee407e624e4978e2f291"}}