{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:SO76PEDCYF7C2AJR32CVFSZNOP","short_pith_number":"pith:SO76PEDC","canonical_record":{"source":{"id":"2304.09871","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-04-19T06:15:11Z","cross_cats_sorted":["cs.AI","math.OC"],"title_canon_sha256":"6553f380d521f4433f5c6b03ae6394880b96859825918bcbb65930d100ff853c","abstract_canon_sha256":"a21c0b5d276017e7c3570540e757eeedc76ce8b44705c1c0294914c1550602ac"},"schema_version":"1.0"},"canonical_sha256":"93bfe79062c17e2d0131de8552cb2d73d50766b387be15adbc66e345222c17e5","source":{"kind":"arxiv","id":"2304.09871","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2304.09871","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"arxiv_version","alias_value":"2304.09871v2","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.09871","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"pith_short_12","alias_value":"SO76PEDCYF7C","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"pith_short_16","alias_value":"SO76PEDCYF7C2AJR","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"pith_short_8","alias_value":"SO76PEDC","created_at":"2026-07-05T06:04:14Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:SO76PEDCYF7C2AJR32CVFSZNOP","target":"record","payload":{"canonical_record":{"source":{"id":"2304.09871","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-04-19T06:15:11Z","cross_cats_sorted":["cs.AI","math.OC"],"title_canon_sha256":"6553f380d521f4433f5c6b03ae6394880b96859825918bcbb65930d100ff853c","abstract_canon_sha256":"a21c0b5d276017e7c3570540e757eeedc76ce8b44705c1c0294914c1550602ac"},"schema_version":"1.0"},"canonical_sha256":"93bfe79062c17e2d0131de8552cb2d73d50766b387be15adbc66e345222c17e5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:04:14.377568Z","signature_b64":"d5D9FPu4l9KIfs7uNfJVVsa/6cgniRvcBkWYI2sS+4TpPHzst0G3vIgXiZ/7wh7nm2n8EY20b7/ycz0cxxe0BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"93bfe79062c17e2d0131de8552cb2d73d50766b387be15adbc66e345222c17e5","last_reissued_at":"2026-07-05T06:04:14.377109Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:04:14.377109Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2304.09871","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":"2026-07-05T06:04:14Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"I2KuuR/NeC8JjWcqCiOkViGZiXPv+EiM9FqO9hEf3RAm2dvq4/VtujME4h2G4T2Q5hoxlcv6lFFGUQWCWFlbAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-10T10:58:42.605425Z"},"content_sha256":"383e38eaec2a4f09d1780b1e37b4aed7d49fcb9ac910507ef08bd280b13cee50","schema_version":"1.0","event_id":"sha256:383e38eaec2a4f09d1780b1e37b4aed7d49fcb9ac910507ef08bd280b13cee50"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:SO76PEDCYF7C2AJR32CVFSZNOP","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"A Theory on Adam Instability in Large-Scale Machine Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","math.OC"],"primary_cat":"cs.LG","authors_text":"Andrew Poulton, Binh Tang, David Esiobu, Diana Liskovich, Igor Molybog, Melanie Kambadur, Moya Chen, Naman Goyal, Peter Albert, Punit Singh Koura, Puxin Xu, Ruan Silva, Sharan Narang, Stephen Roller, Susan Zhang, Yuchen Zhang, Zachary DeVito","submitted_at":"2023-04-19T06:15:11Z","abstract_excerpt":"We present a theory for the previously unexplained divergent behavior noticed in the training of large language models. We argue that the phenomenon is an artifact of the dominant optimization algorithm used for training, called Adam. We observe that Adam can enter a state in which the parameter update vector has a relatively large norm and is essentially uncorrelated with the direction of descent on the training loss landscape, leading to divergence. This artifact is more likely to be observed in the training of a deep model with a large batch size, which is the typical setting of large-scale"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2304.09871","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/2304.09871/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":"2026-07-05T06:04:14Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"STtDJhcF4n90fXZUxp0+PNdDcpPLwFFcWtq1r/7rdp7lcE9NLuM8IX5Xb4YxPWHfUlff5k5cbwvHDwDyh2pnBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-10T10:58:42.605820Z"},"content_sha256":"95d3994f0ef0c7df8eec807cda01d4eaa64360ab7fae877f74a5f6dab3526d2a","schema_version":"1.0","event_id":"sha256:95d3994f0ef0c7df8eec807cda01d4eaa64360ab7fae877f74a5f6dab3526d2a"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/SO76PEDCYF7C2AJR32CVFSZNOP/bundle.json","state_url":"https://pith.science/pith/SO76PEDCYF7C2AJR32CVFSZNOP/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/SO76PEDCYF7C2AJR32CVFSZNOP/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":"2026-07-10T10:58:42Z","links":{"resolver":"https://pith.science/pith/SO76PEDCYF7C2AJR32CVFSZNOP","bundle":"https://pith.science/pith/SO76PEDCYF7C2AJR32CVFSZNOP/bundle.json","state":"https://pith.science/pith/SO76PEDCYF7C2AJR32CVFSZNOP/state.json","well_known_bundle":"https://pith.science/.well-known/pith/SO76PEDCYF7C2AJR32CVFSZNOP/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:SO76PEDCYF7C2AJR32CVFSZNOP","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":"a21c0b5d276017e7c3570540e757eeedc76ce8b44705c1c0294914c1550602ac","cross_cats_sorted":["cs.AI","math.OC"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-04-19T06:15:11Z","title_canon_sha256":"6553f380d521f4433f5c6b03ae6394880b96859825918bcbb65930d100ff853c"},"schema_version":"1.0","source":{"id":"2304.09871","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2304.09871","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"arxiv_version","alias_value":"2304.09871v2","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.09871","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"pith_short_12","alias_value":"SO76PEDCYF7C","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"pith_short_16","alias_value":"SO76PEDCYF7C2AJR","created_at":"2026-07-05T06:04:14Z"},{"alias_kind":"pith_short_8","alias_value":"SO76PEDC","created_at":"2026-07-05T06:04:14Z"}],"graph_snapshots":[{"event_id":"sha256:95d3994f0ef0c7df8eec807cda01d4eaa64360ab7fae877f74a5f6dab3526d2a","target":"graph","created_at":"2026-07-05T06:04:14Z","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/2304.09871/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We present a theory for the previously unexplained divergent behavior noticed in the training of large language models. We argue that the phenomenon is an artifact of the dominant optimization algorithm used for training, called Adam. We observe that Adam can enter a state in which the parameter update vector has a relatively large norm and is essentially uncorrelated with the direction of descent on the training loss landscape, leading to divergence. This artifact is more likely to be observed in the training of a deep model with a large batch size, which is the typical setting of large-scale","authors_text":"Andrew Poulton, Binh Tang, David Esiobu, Diana Liskovich, Igor Molybog, Melanie Kambadur, Moya Chen, Naman Goyal, Peter Albert, Punit Singh Koura, Puxin Xu, Ruan Silva, Sharan Narang, Stephen Roller, Susan Zhang, Yuchen Zhang, Zachary DeVito","cross_cats":["cs.AI","math.OC"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-04-19T06:15:11Z","title":"A Theory on Adam Instability in Large-Scale Machine Learning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2304.09871","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":"sha256:383e38eaec2a4f09d1780b1e37b4aed7d49fcb9ac910507ef08bd280b13cee50","target":"record","created_at":"2026-07-05T06:04:14Z","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":"a21c0b5d276017e7c3570540e757eeedc76ce8b44705c1c0294914c1550602ac","cross_cats_sorted":["cs.AI","math.OC"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-04-19T06:15:11Z","title_canon_sha256":"6553f380d521f4433f5c6b03ae6394880b96859825918bcbb65930d100ff853c"},"schema_version":"1.0","source":{"id":"2304.09871","kind":"arxiv","version":2}},"canonical_sha256":"93bfe79062c17e2d0131de8552cb2d73d50766b387be15adbc66e345222c17e5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"93bfe79062c17e2d0131de8552cb2d73d50766b387be15adbc66e345222c17e5","first_computed_at":"2026-07-05T06:04:14.377109Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T06:04:14.377109Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"d5D9FPu4l9KIfs7uNfJVVsa/6cgniRvcBkWYI2sS+4TpPHzst0G3vIgXiZ/7wh7nm2n8EY20b7/ycz0cxxe0BQ==","signature_status":"signed_v1","signed_at":"2026-07-05T06:04:14.377568Z","signed_message":"canonical_sha256_bytes"},"source_id":"2304.09871","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:383e38eaec2a4f09d1780b1e37b4aed7d49fcb9ac910507ef08bd280b13cee50","sha256:95d3994f0ef0c7df8eec807cda01d4eaa64360ab7fae877f74a5f6dab3526d2a"],"state_sha256":"504544bcae00445bcd6f0b64c20c93f3eb4e44bac724da1482eb70a7ac08a42e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JpZh0QEMgEypsIzAi3Rp5z1DPpemFQRuwLPrTur8Go/5yDHSAfU6u4xStEZO/zQzfmPeJoWU33pZJsxcmSrfDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-10T10:58:42.607933Z","bundle_sha256":"34578c61e5ae6cecb45a6e1d26bc3014c1f64548e9fd38c904c2b060cf910448"}}