{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:FY5OCOPMZSZP3HFPYC4MITXXBZ","short_pith_number":"pith:FY5OCOPM","canonical_record":{"source":{"id":"2504.02507","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-03T11:41:55Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"a99ed1cba59c9fa3765dab3bdc6023dd2a21ad4bf6e8c546476c70f31ddc7883","abstract_canon_sha256":"6832fd59a78815011b27acc641bdd88db700fafd5e345ff2c88bff0de0cbdde8"},"schema_version":"1.0"},"canonical_sha256":"2e3ae139ecccb2fd9cafc0b8c44ef70e49cb8610744a5396a90b47e3ed84db92","source":{"kind":"arxiv","id":"2504.02507","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.02507","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"arxiv_version","alias_value":"2504.02507v1","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.02507","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"pith_short_12","alias_value":"FY5OCOPMZSZP","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"pith_short_16","alias_value":"FY5OCOPMZSZP3HFP","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"pith_short_8","alias_value":"FY5OCOPM","created_at":"2026-07-05T10:43:58Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:FY5OCOPMZSZP3HFPYC4MITXXBZ","target":"record","payload":{"canonical_record":{"source":{"id":"2504.02507","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-03T11:41:55Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"a99ed1cba59c9fa3765dab3bdc6023dd2a21ad4bf6e8c546476c70f31ddc7883","abstract_canon_sha256":"6832fd59a78815011b27acc641bdd88db700fafd5e345ff2c88bff0de0cbdde8"},"schema_version":"1.0"},"canonical_sha256":"2e3ae139ecccb2fd9cafc0b8c44ef70e49cb8610744a5396a90b47e3ed84db92","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:43:58.228237Z","signature_b64":"yL3l3XLsUtFDjJp9Q4NxE4zk9E7KjDg6gW1nU7REaxHPiRZVsZ+IRg3zTc6ybRpK+Gm8QpRxlzciDOZMhQiMAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2e3ae139ecccb2fd9cafc0b8c44ef70e49cb8610744a5396a90b47e3ed84db92","last_reissued_at":"2026-07-05T10:43:58.227738Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:43:58.227738Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2504.02507","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:43:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Kd4Z0We3/nUYgjA2exQiMsi1suy1SqrR392Hvief7MkKvZYBpF4imPOYpfIa6brjetm34ocze/JdocSqJr25BQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T22:27:00.269758Z"},"content_sha256":"ae8c7803fde0ce73bc7f45d757e648f9368f558cefb3de446aa023aefedd7dec","schema_version":"1.0","event_id":"sha256:ae8c7803fde0ce73bc7f45d757e648f9368f558cefb3de446aa023aefedd7dec"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:FY5OCOPMZSZP3HFPYC4MITXXBZ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"ZClip: Adaptive Spike Mitigation for LLM Pre-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Abhay Kumar, Fabian G\\\"ura, Louis Owen, Nilabhra Roy Chowdhury","submitted_at":"2025-04-03T11:41:55Z","abstract_excerpt":"Training large language models (LLMs) presents numerous challenges, including gradient instability and loss spikes. These phenomena can lead to catastrophic divergence, requiring costly checkpoint restoration and data batch skipping. Traditional gradient clipping techniques, such as constant or norm-based methods, fail to address these issues effectively due to their reliance on fixed thresholds or heuristics, leading to inefficient learning and requiring frequent manual intervention. In this work, we propose ZClip, an adaptive gradient clipping algorithm that dynamically adjusts the clipping "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.02507","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/2504.02507/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":"2026-07-05T10:43:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JK8lLmHvuYLnlfNhuqVMQwp+5XUW/TGqU/dSPycLn3Waha4K59T0ulPowBmEl6eKbj08+dStH6AT6idHn0KiAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T22:27:00.270264Z"},"content_sha256":"9de03e68a8e47c2b310551552a08b45ff0944050de207a7a17c2309c0b78a02f","schema_version":"1.0","event_id":"sha256:9de03e68a8e47c2b310551552a08b45ff0944050de207a7a17c2309c0b78a02f"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/FY5OCOPMZSZP3HFPYC4MITXXBZ/bundle.json","state_url":"https://pith.science/pith/FY5OCOPMZSZP3HFPYC4MITXXBZ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/FY5OCOPMZSZP3HFPYC4MITXXBZ/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":"2026-08-03T22:27:00Z","links":{"resolver":"https://pith.science/pith/FY5OCOPMZSZP3HFPYC4MITXXBZ","bundle":"https://pith.science/pith/FY5OCOPMZSZP3HFPYC4MITXXBZ/bundle.json","state":"https://pith.science/pith/FY5OCOPMZSZP3HFPYC4MITXXBZ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/FY5OCOPMZSZP3HFPYC4MITXXBZ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:FY5OCOPMZSZP3HFPYC4MITXXBZ","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":"6832fd59a78815011b27acc641bdd88db700fafd5e345ff2c88bff0de0cbdde8","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-03T11:41:55Z","title_canon_sha256":"a99ed1cba59c9fa3765dab3bdc6023dd2a21ad4bf6e8c546476c70f31ddc7883"},"schema_version":"1.0","source":{"id":"2504.02507","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.02507","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"arxiv_version","alias_value":"2504.02507v1","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.02507","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"pith_short_12","alias_value":"FY5OCOPMZSZP","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"pith_short_16","alias_value":"FY5OCOPMZSZP3HFP","created_at":"2026-07-05T10:43:58Z"},{"alias_kind":"pith_short_8","alias_value":"FY5OCOPM","created_at":"2026-07-05T10:43:58Z"}],"graph_snapshots":[{"event_id":"sha256:9de03e68a8e47c2b310551552a08b45ff0944050de207a7a17c2309c0b78a02f","target":"graph","created_at":"2026-07-05T10:43:58Z","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/2504.02507/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Training large language models (LLMs) presents numerous challenges, including gradient instability and loss spikes. These phenomena can lead to catastrophic divergence, requiring costly checkpoint restoration and data batch skipping. Traditional gradient clipping techniques, such as constant or norm-based methods, fail to address these issues effectively due to their reliance on fixed thresholds or heuristics, leading to inefficient learning and requiring frequent manual intervention. In this work, we propose ZClip, an adaptive gradient clipping algorithm that dynamically adjusts the clipping ","authors_text":"Abhay Kumar, Fabian G\\\"ura, Louis Owen, Nilabhra Roy Chowdhury","cross_cats":["cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-03T11:41:55Z","title":"ZClip: Adaptive Spike Mitigation for LLM Pre-Training"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.02507","kind":"arxiv","version":1},"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:ae8c7803fde0ce73bc7f45d757e648f9368f558cefb3de446aa023aefedd7dec","target":"record","created_at":"2026-07-05T10:43:58Z","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":"6832fd59a78815011b27acc641bdd88db700fafd5e345ff2c88bff0de0cbdde8","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-03T11:41:55Z","title_canon_sha256":"a99ed1cba59c9fa3765dab3bdc6023dd2a21ad4bf6e8c546476c70f31ddc7883"},"schema_version":"1.0","source":{"id":"2504.02507","kind":"arxiv","version":1}},"canonical_sha256":"2e3ae139ecccb2fd9cafc0b8c44ef70e49cb8610744a5396a90b47e3ed84db92","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2e3ae139ecccb2fd9cafc0b8c44ef70e49cb8610744a5396a90b47e3ed84db92","first_computed_at":"2026-07-05T10:43:58.227738Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:43:58.227738Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"yL3l3XLsUtFDjJp9Q4NxE4zk9E7KjDg6gW1nU7REaxHPiRZVsZ+IRg3zTc6ybRpK+Gm8QpRxlzciDOZMhQiMAw==","signature_status":"signed_v1","signed_at":"2026-07-05T10:43:58.228237Z","signed_message":"canonical_sha256_bytes"},"source_id":"2504.02507","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ae8c7803fde0ce73bc7f45d757e648f9368f558cefb3de446aa023aefedd7dec","sha256:9de03e68a8e47c2b310551552a08b45ff0944050de207a7a17c2309c0b78a02f"],"state_sha256":"128f386594948f25e9c75732f9702c5df0378eeebb0eb34fc646c1bc168ec1a5"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1mAqAT3CjCQTHPXjC4zZ1vYF6dICXIGI3U+zZwz/zCVgMFE7+MNXPAEd53xAU6Pmg6J7d5DEtUv1QDEnQoflCg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-03T22:27:00.274018Z","bundle_sha256":"a0392b7ef85998c51130b815d0cce8444077d25ac5dc0129a5dac7acf4130222"}}