{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:RH672XDIC7QVDWOHB45MZBGHKP","short_pith_number":"pith:RH672XDI","canonical_record":{"source":{"id":"2406.04443","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T18:49:10Z","cross_cats_sorted":["math.OC"],"title_canon_sha256":"abb18bb522547cae3408a07453c7c2221da059a73cf5421ee6e99226cbf56c65","abstract_canon_sha256":"36d15ce3e54aa32d8b46b901d551f2ccaf0341f8e179dd5a3e7b1c3be85aeb1c"},"schema_version":"1.0"},"canonical_sha256":"89fdfd5c6817e151d9c70f3acc84c753e54d4b0c193cf59c01871dad9ed0a940","source":{"kind":"arxiv","id":"2406.04443","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.04443","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"arxiv_version","alias_value":"2406.04443v3","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.04443","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"pith_short_12","alias_value":"RH672XDIC7QV","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"pith_short_16","alias_value":"RH672XDIC7QVDWOH","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"pith_short_8","alias_value":"RH672XDI","created_at":"2026-07-05T11:53:40Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:RH672XDIC7QVDWOHB45MZBGHKP","target":"record","payload":{"canonical_record":{"source":{"id":"2406.04443","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T18:49:10Z","cross_cats_sorted":["math.OC"],"title_canon_sha256":"abb18bb522547cae3408a07453c7c2221da059a73cf5421ee6e99226cbf56c65","abstract_canon_sha256":"36d15ce3e54aa32d8b46b901d551f2ccaf0341f8e179dd5a3e7b1c3be85aeb1c"},"schema_version":"1.0"},"canonical_sha256":"89fdfd5c6817e151d9c70f3acc84c753e54d4b0c193cf59c01871dad9ed0a940","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:53:40.934181Z","signature_b64":"IswWaJ8bCj6QDQTOUDMDe8Gb61Cq+3zWjLv7YgFoT6DP06chu/JoYIpIE7c9WstSSHaHB3YRJFkkU62vT1C5DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"89fdfd5c6817e151d9c70f3acc84c753e54d4b0c193cf59c01871dad9ed0a940","last_reissued_at":"2026-07-05T11:53:40.933689Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:53:40.933689Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.04443","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:53:40Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Tncr0hKtw33ODIeVSj3y/hrVVHjtzPDnVqNT1/qishhz7AUHBTFIGvWey7EHNJXLQdsSf6Fj6m0Bdc8eX5TICg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T00:15:43.305481Z"},"content_sha256":"c5a1fd536109378bab6337654e6b865897b5cb44886875d149d5e30911634a41","schema_version":"1.0","event_id":"sha256:c5a1fd536109378bab6337654e6b865897b5cb44886875d149d5e30911634a41"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:RH672XDIC7QVDWOHB45MZBGHKP","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Clipping Improves Adam-Norm and AdaGrad-Norm when the Noise Is Heavy-Tailed","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Aleksandr Beznosikov, Alexander Gasnikov, Andrei Semenov, Eduard Gorbunov, Martin Tak\\'a\\v{c}, Samuel Horv\\'ath, Savelii Chezhegov, Yaroslav Klyukin","submitted_at":"2024-06-06T18:49:10Z","abstract_excerpt":"Methods with adaptive stepsizes, such as AdaGrad and Adam, are essential for training modern Deep Learning models, especially Large Language Models. Typically, the noise in the stochastic gradients is heavy-tailed for the later ones. Gradient clipping provably helps to achieve good high-probability convergence for such noises. However, despite the similarity between AdaGrad/Adam and Clip-SGD, the current understanding of the high-probability convergence of AdaGrad/Adam-type methods is limited in this case. In this work, we prove that AdaGrad/Adam (and their delayed version) can have provably b"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.04443","kind":"arxiv","version":3},"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/2406.04443/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":"2026-07-05T11:53:40Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"UhtyIYMrQX8Y1m6NrDDkMfRpWEMh/H/Pxee4U6ZAsaOaILV5DfENXS49UtGsJtssd4FdGSFR6BV02kIdlusBBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T00:15:43.305998Z"},"content_sha256":"2919e0ee283c34e0a517b35b4b250c3679c6774560e40317c46d5a4b026395b7","schema_version":"1.0","event_id":"sha256:2919e0ee283c34e0a517b35b4b250c3679c6774560e40317c46d5a4b026395b7"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/RH672XDIC7QVDWOHB45MZBGHKP/bundle.json","state_url":"https://pith.science/pith/RH672XDIC7QVDWOHB45MZBGHKP/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/RH672XDIC7QVDWOHB45MZBGHKP/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":"2026-08-09T00:15:43Z","links":{"resolver":"https://pith.science/pith/RH672XDIC7QVDWOHB45MZBGHKP","bundle":"https://pith.science/pith/RH672XDIC7QVDWOHB45MZBGHKP/bundle.json","state":"https://pith.science/pith/RH672XDIC7QVDWOHB45MZBGHKP/state.json","well_known_bundle":"https://pith.science/.well-known/pith/RH672XDIC7QVDWOHB45MZBGHKP/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:RH672XDIC7QVDWOHB45MZBGHKP","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":"36d15ce3e54aa32d8b46b901d551f2ccaf0341f8e179dd5a3e7b1c3be85aeb1c","cross_cats_sorted":["math.OC"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T18:49:10Z","title_canon_sha256":"abb18bb522547cae3408a07453c7c2221da059a73cf5421ee6e99226cbf56c65"},"schema_version":"1.0","source":{"id":"2406.04443","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.04443","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"arxiv_version","alias_value":"2406.04443v3","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.04443","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"pith_short_12","alias_value":"RH672XDIC7QV","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"pith_short_16","alias_value":"RH672XDIC7QVDWOH","created_at":"2026-07-05T11:53:40Z"},{"alias_kind":"pith_short_8","alias_value":"RH672XDI","created_at":"2026-07-05T11:53:40Z"}],"graph_snapshots":[{"event_id":"sha256:2919e0ee283c34e0a517b35b4b250c3679c6774560e40317c46d5a4b026395b7","target":"graph","created_at":"2026-07-05T11:53:40Z","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/2406.04443/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Methods with adaptive stepsizes, such as AdaGrad and Adam, are essential for training modern Deep Learning models, especially Large Language Models. Typically, the noise in the stochastic gradients is heavy-tailed for the later ones. Gradient clipping provably helps to achieve good high-probability convergence for such noises. However, despite the similarity between AdaGrad/Adam and Clip-SGD, the current understanding of the high-probability convergence of AdaGrad/Adam-type methods is limited in this case. In this work, we prove that AdaGrad/Adam (and their delayed version) can have provably b","authors_text":"Aleksandr Beznosikov, Alexander Gasnikov, Andrei Semenov, Eduard Gorbunov, Martin Tak\\'a\\v{c}, Samuel Horv\\'ath, Savelii Chezhegov, Yaroslav Klyukin","cross_cats":["math.OC"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T18:49:10Z","title":"Clipping Improves Adam-Norm and AdaGrad-Norm when the Noise Is Heavy-Tailed"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.04443","kind":"arxiv","version":3},"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:c5a1fd536109378bab6337654e6b865897b5cb44886875d149d5e30911634a41","target":"record","created_at":"2026-07-05T11:53:40Z","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":"36d15ce3e54aa32d8b46b901d551f2ccaf0341f8e179dd5a3e7b1c3be85aeb1c","cross_cats_sorted":["math.OC"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T18:49:10Z","title_canon_sha256":"abb18bb522547cae3408a07453c7c2221da059a73cf5421ee6e99226cbf56c65"},"schema_version":"1.0","source":{"id":"2406.04443","kind":"arxiv","version":3}},"canonical_sha256":"89fdfd5c6817e151d9c70f3acc84c753e54d4b0c193cf59c01871dad9ed0a940","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"89fdfd5c6817e151d9c70f3acc84c753e54d4b0c193cf59c01871dad9ed0a940","first_computed_at":"2026-07-05T11:53:40.933689Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:53:40.933689Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"IswWaJ8bCj6QDQTOUDMDe8Gb61Cq+3zWjLv7YgFoT6DP06chu/JoYIpIE7c9WstSSHaHB3YRJFkkU62vT1C5DQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:53:40.934181Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.04443","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c5a1fd536109378bab6337654e6b865897b5cb44886875d149d5e30911634a41","sha256:2919e0ee283c34e0a517b35b4b250c3679c6774560e40317c46d5a4b026395b7"],"state_sha256":"e51fccbbe92c56b68281cb2ed8e1f5f7e1b797119a7011c2019ecd3f9aa26a00"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XmdPTDInT4UWmTWYqNzX2M50tqMEZo2NDRTTfi2rSV/Ef9dj43QmMh2S8IAdrtl2F7WE3s48aY9ytZQmvBjRDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T00:15:43.309372Z","bundle_sha256":"0f185db4068a38ae7a45456b294e03badd332a1fe6c8441faab12c577f1cb7b7"}}