{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XW7ILGAJFV6WLUSXODZCPQB5GZ","short_pith_number":"pith:XW7ILGAJ","schema_version":"1.0","canonical_sha256":"bdbe8598092d7d65d25770f227c03d364bc15e476fe388da6b837ba88731465b","source":{"kind":"arxiv","id":"2606.30634","version":1},"attestation_state":"computed","paper":{"title":"One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Egor Petrov, Mikhail Khrushchev, Nursultan Abdullaev, Philip Zmushko, Samuel Horv\\'ath","submitted_at":"2026-06-29T17:57:50Z","abstract_excerpt":"Modern large-scale LLM pretraining benefits from utilizing Pipeline Parallelism; however, synchronous implementations leave GPUs idle during pipeline bubbles, wasting computational resources. Asynchronous Pipeline Parallelism eliminates these bubbles, maximizing throughput at the cost of gradient staleness. Among asynchronous schedules, PipeDream-2BW is particularly appealing: unlike the original PipeDream schedule, it ensures a constant one-step gradient delay regardless of pipeline depth. However, its adoption remains limited due to the common belief that optimizing under staleness is fundam"},"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":"2606.30634","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-29T17:57:50Z","cross_cats_sorted":[],"title_canon_sha256":"4735b36e21ec52eed969351560ae9d22a81a3820a0cf3218e014c10062c75e32","abstract_canon_sha256":"7bea4c2369e3e1f900e59d95c643b2a9b290cf4852f0253289d3781cecde4374"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T02:18:23.743481Z","signature_b64":"RpoiVYXu4wUkGoOol81w8H0AtTwyfh0dudemhWSJqtVRBLxKQLSXOKOUEUbhDYy00AqUzFJ+Vsab50V9o7BiBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bdbe8598092d7d65d25770f227c03d364bc15e476fe388da6b837ba88731465b","last_reissued_at":"2026-06-30T02:18:23.742770Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T02:18:23.742770Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Egor Petrov, Mikhail Khrushchev, Nursultan Abdullaev, Philip Zmushko, Samuel Horv\\'ath","submitted_at":"2026-06-29T17:57:50Z","abstract_excerpt":"Modern large-scale LLM pretraining benefits from utilizing Pipeline Parallelism; however, synchronous implementations leave GPUs idle during pipeline bubbles, wasting computational resources. Asynchronous Pipeline Parallelism eliminates these bubbles, maximizing throughput at the cost of gradient staleness. Among asynchronous schedules, PipeDream-2BW is particularly appealing: unlike the original PipeDream schedule, it ensures a constant one-step gradient delay regardless of pipeline depth. However, its adoption remains limited due to the common belief that optimizing under staleness is fundam"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.30634","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/2606.30634/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":"2606.30634","created_at":"2026-06-30T02:18:23.742863+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.30634v1","created_at":"2026-06-30T02:18:23.742863+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.30634","created_at":"2026-06-30T02:18:23.742863+00:00"},{"alias_kind":"pith_short_12","alias_value":"XW7ILGAJFV6W","created_at":"2026-06-30T02:18:23.742863+00:00"},{"alias_kind":"pith_short_16","alias_value":"XW7ILGAJFV6WLUSX","created_at":"2026-06-30T02:18:23.742863+00:00"},{"alias_kind":"pith_short_8","alias_value":"XW7ILGAJ","created_at":"2026-06-30T02:18:23.742863+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/XW7ILGAJFV6WLUSXODZCPQB5GZ","json":"https://pith.science/pith/XW7ILGAJFV6WLUSXODZCPQB5GZ.json","graph_json":"https://pith.science/api/pith-number/XW7ILGAJFV6WLUSXODZCPQB5GZ/graph.json","events_json":"https://pith.science/api/pith-number/XW7ILGAJFV6WLUSXODZCPQB5GZ/events.json","paper":"https://pith.science/paper/XW7ILGAJ"},"agent_actions":{"view_html":"https://pith.science/pith/XW7ILGAJFV6WLUSXODZCPQB5GZ","download_json":"https://pith.science/pith/XW7ILGAJFV6WLUSXODZCPQB5GZ.json","view_paper":"https://pith.science/paper/XW7ILGAJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.30634&json=true","fetch_graph":"https://pith.science/api/pith-number/XW7ILGAJFV6WLUSXODZCPQB5GZ/graph.json","fetch_events":"https://pith.science/api/pith-number/XW7ILGAJFV6WLUSXODZCPQB5GZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XW7ILGAJFV6WLUSXODZCPQB5GZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XW7ILGAJFV6WLUSXODZCPQB5GZ/action/storage_attestation","attest_author":"https://pith.science/pith/XW7ILGAJFV6WLUSXODZCPQB5GZ/action/author_attestation","sign_citation":"https://pith.science/pith/XW7ILGAJFV6WLUSXODZCPQB5GZ/action/citation_signature","submit_replication":"https://pith.science/pith/XW7ILGAJFV6WLUSXODZCPQB5GZ/action/replication_record"}},"created_at":"2026-06-30T02:18:23.742863+00:00","updated_at":"2026-06-30T02:18:23.742863+00:00"}