{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:H74EJCWVOLINB2RYVQCEQDANX2","short_pith_number":"pith:H74EJCWV","canonical_record":{"source":{"id":"2406.09723","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-14T05:17:39Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"467c6e4a36fdc687de8b6127027df9642a3543d76f64d69ba918ba6e4ff55606","abstract_canon_sha256":"440bda18308009208c38595130d7e96fe6594daf53f0b211348de1d0fae9e0b3"},"schema_version":"1.0"},"canonical_sha256":"3ff8448ad572d0d0ea38ac04480c0dbe80e223f90eca95ec22cd894a17d6f9d4","source":{"kind":"arxiv","id":"2406.09723","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.09723","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"arxiv_version","alias_value":"2406.09723v1","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.09723","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"pith_short_12","alias_value":"H74EJCWVOLIN","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"pith_short_16","alias_value":"H74EJCWVOLINB2RY","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"pith_short_8","alias_value":"H74EJCWV","created_at":"2026-07-05T08:31:58Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:H74EJCWVOLINB2RYVQCEQDANX2","target":"record","payload":{"canonical_record":{"source":{"id":"2406.09723","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-14T05:17:39Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"467c6e4a36fdc687de8b6127027df9642a3543d76f64d69ba918ba6e4ff55606","abstract_canon_sha256":"440bda18308009208c38595130d7e96fe6594daf53f0b211348de1d0fae9e0b3"},"schema_version":"1.0"},"canonical_sha256":"3ff8448ad572d0d0ea38ac04480c0dbe80e223f90eca95ec22cd894a17d6f9d4","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:31:58.848671Z","signature_b64":"7NvI/XsblpXsy73T0UrKWXpZynvy8kpaFRWR1OHTUaQ9mdk6jSnIdxAMAOEpHrux84KgGe7k6NPICyFfXcPKBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3ff8448ad572d0d0ea38ac04480c0dbe80e223f90eca95ec22cd894a17d6f9d4","last_reissued_at":"2026-07-05T08:31:58.848184Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:31:58.848184Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.09723","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":"2026-07-05T08:31:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+AsMt32uL59AncBfpWYzL61d3PkFhpOlZwff+D+Vwc77o9rP6ret7aOWAC3VEIMq0KTgcgGH1hqw0qFjk/gFAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T05:46:17.873748Z"},"content_sha256":"1df31b99a93edbe2eb5368cd5812657398838844f5062c1d101a1e1972573ce9","schema_version":"1.0","event_id":"sha256:1df31b99a93edbe2eb5368cd5812657398838844f5062c1d101a1e1972573ce9"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:H74EJCWVOLINB2RYVQCEQDANX2","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"When Will Gradient Regularization Be Harmful?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hao Zhang, Xiuyuan Hu, Yang Zhao","submitted_at":"2024-06-14T05:17:39Z","abstract_excerpt":"Gradient regularization (GR), which aims to penalize the gradient norm atop the loss function, has shown promising results in training modern over-parameterized deep neural networks. However, can we trust this powerful technique? This paper reveals that GR can cause performance degeneration in adaptive optimization scenarios, particularly with learning rate warmup. Our empirical and theoretical analyses suggest this is due to GR inducing instability and divergence in gradient statistics of adaptive optimizers at the initial training stage. Inspired by the warmup heuristic, we propose three GR "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.09723","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/2406.09723/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":"2026-07-05T08:31:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"eDLFK43cHBvvjGyQ92Si7E/pvWdZxmhAS6DagqkcEFMKZP99DPi/qbNMuVQ2MvHjAXmMeFLaRbAk1Qfb/MpTDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T05:46:17.874420Z"},"content_sha256":"633f1e970e7a5322308185ac52899715addc572cbc5a0bc7043f54242d0b8721","schema_version":"1.0","event_id":"sha256:633f1e970e7a5322308185ac52899715addc572cbc5a0bc7043f54242d0b8721"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/H74EJCWVOLINB2RYVQCEQDANX2/bundle.json","state_url":"https://pith.science/pith/H74EJCWVOLINB2RYVQCEQDANX2/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/H74EJCWVOLINB2RYVQCEQDANX2/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":"2026-07-07T05:46:17Z","links":{"resolver":"https://pith.science/pith/H74EJCWVOLINB2RYVQCEQDANX2","bundle":"https://pith.science/pith/H74EJCWVOLINB2RYVQCEQDANX2/bundle.json","state":"https://pith.science/pith/H74EJCWVOLINB2RYVQCEQDANX2/state.json","well_known_bundle":"https://pith.science/.well-known/pith/H74EJCWVOLINB2RYVQCEQDANX2/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:H74EJCWVOLINB2RYVQCEQDANX2","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":"440bda18308009208c38595130d7e96fe6594daf53f0b211348de1d0fae9e0b3","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-14T05:17:39Z","title_canon_sha256":"467c6e4a36fdc687de8b6127027df9642a3543d76f64d69ba918ba6e4ff55606"},"schema_version":"1.0","source":{"id":"2406.09723","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.09723","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"arxiv_version","alias_value":"2406.09723v1","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.09723","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"pith_short_12","alias_value":"H74EJCWVOLIN","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"pith_short_16","alias_value":"H74EJCWVOLINB2RY","created_at":"2026-07-05T08:31:58Z"},{"alias_kind":"pith_short_8","alias_value":"H74EJCWV","created_at":"2026-07-05T08:31:58Z"}],"graph_snapshots":[{"event_id":"sha256:633f1e970e7a5322308185ac52899715addc572cbc5a0bc7043f54242d0b8721","target":"graph","created_at":"2026-07-05T08:31:58Z","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/2406.09723/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Gradient regularization (GR), which aims to penalize the gradient norm atop the loss function, has shown promising results in training modern over-parameterized deep neural networks. However, can we trust this powerful technique? This paper reveals that GR can cause performance degeneration in adaptive optimization scenarios, particularly with learning rate warmup. Our empirical and theoretical analyses suggest this is due to GR inducing instability and divergence in gradient statistics of adaptive optimizers at the initial training stage. Inspired by the warmup heuristic, we propose three GR ","authors_text":"Hao Zhang, Xiuyuan Hu, Yang Zhao","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-14T05:17:39Z","title":"When Will Gradient Regularization Be Harmful?"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.09723","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":"sha256:1df31b99a93edbe2eb5368cd5812657398838844f5062c1d101a1e1972573ce9","target":"record","created_at":"2026-07-05T08:31:58Z","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":"440bda18308009208c38595130d7e96fe6594daf53f0b211348de1d0fae9e0b3","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-14T05:17:39Z","title_canon_sha256":"467c6e4a36fdc687de8b6127027df9642a3543d76f64d69ba918ba6e4ff55606"},"schema_version":"1.0","source":{"id":"2406.09723","kind":"arxiv","version":1}},"canonical_sha256":"3ff8448ad572d0d0ea38ac04480c0dbe80e223f90eca95ec22cd894a17d6f9d4","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3ff8448ad572d0d0ea38ac04480c0dbe80e223f90eca95ec22cd894a17d6f9d4","first_computed_at":"2026-07-05T08:31:58.848184Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:31:58.848184Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"7NvI/XsblpXsy73T0UrKWXpZynvy8kpaFRWR1OHTUaQ9mdk6jSnIdxAMAOEpHrux84KgGe7k6NPICyFfXcPKBQ==","signature_status":"signed_v1","signed_at":"2026-07-05T08:31:58.848671Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.09723","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1df31b99a93edbe2eb5368cd5812657398838844f5062c1d101a1e1972573ce9","sha256:633f1e970e7a5322308185ac52899715addc572cbc5a0bc7043f54242d0b8721"],"state_sha256":"a67998a5e5df304b7495684a4bf2c4a194de0d3e9b60e7d2443567111cb31e11"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"TRiFaols3+FLcydh0X+fZZ16czE4hG+Gx+mXfsgt0t6B13grzel1BQb7iy2UcmrYQpGE9eATyVp2VyI+x6GCCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-07T05:46:17.877559Z","bundle_sha256":"6053df39742d5132256480ca55f198f707bdd865de06f9c983d651ee4ff94d93"}}