{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:HUQINWJ4CJ4QBRRKROF4WCZYHO","short_pith_number":"pith:HUQINWJ4","schema_version":"1.0","canonical_sha256":"3d2086d93c127900c62a8b8bcb0b383b8a15bbd4b315323f49863accd2d83526","source":{"kind":"arxiv","id":"2310.19215","version":1},"attestation_state":"computed","paper":{"title":"On the accuracy and efficiency of group-wise clipping in differentially private optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CC","cs.CR"],"primary_cat":"cs.LG","authors_text":"George Karypis, Ruixuan Liu, Sheng Zha, Yu-Xiang Wang, Zhiqi Bu","submitted_at":"2023-10-30T01:01:15Z","abstract_excerpt":"Recent advances have substantially improved the accuracy, memory cost, and training speed of differentially private (DP) deep learning, especially on large vision and language models with millions to billions of parameters. In this work, we thoroughly study the per-sample gradient clipping style, a key component in DP optimization. We show that different clipping styles have the same time complexity but instantiate an accuracy-memory trade-off: while the all-layer clipping (of coarse granularity) is the most prevalent and usually gives the best accuracy, it incurs heavier memory cost compared "},"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":"2310.19215","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-10-30T01:01:15Z","cross_cats_sorted":["cs.CC","cs.CR"],"title_canon_sha256":"443a9a1694ff21b2411d6b913d900bd889374d688a82bb9e62fc715559cb6a87","abstract_canon_sha256":"4d5334ada1fe760643fe1469156fd84e78d5d8f0f9058e72d4914985ce82ecb8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:06:47.765007Z","signature_b64":"yIgaWVOJY7hX7q/33uT/l7HWnkYGbfZVkn0VYnhIAVWVnutUbIR3zEKvmMud5c87bYe80pDdBuUU8Omen/VRBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3d2086d93c127900c62a8b8bcb0b383b8a15bbd4b315323f49863accd2d83526","last_reissued_at":"2026-07-05T07:06:47.764524Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:06:47.764524Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the accuracy and efficiency of group-wise clipping in differentially private optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CC","cs.CR"],"primary_cat":"cs.LG","authors_text":"George Karypis, Ruixuan Liu, Sheng Zha, Yu-Xiang Wang, Zhiqi Bu","submitted_at":"2023-10-30T01:01:15Z","abstract_excerpt":"Recent advances have substantially improved the accuracy, memory cost, and training speed of differentially private (DP) deep learning, especially on large vision and language models with millions to billions of parameters. In this work, we thoroughly study the per-sample gradient clipping style, a key component in DP optimization. We show that different clipping styles have the same time complexity but instantiate an accuracy-memory trade-off: while the all-layer clipping (of coarse granularity) is the most prevalent and usually gives the best accuracy, it incurs heavier memory cost compared "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.19215","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/2310.19215/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":"2310.19215","created_at":"2026-07-05T07:06:47.764583+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.19215v1","created_at":"2026-07-05T07:06:47.764583+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.19215","created_at":"2026-07-05T07:06:47.764583+00:00"},{"alias_kind":"pith_short_12","alias_value":"HUQINWJ4CJ4Q","created_at":"2026-07-05T07:06:47.764583+00:00"},{"alias_kind":"pith_short_16","alias_value":"HUQINWJ4CJ4QBRRK","created_at":"2026-07-05T07:06:47.764583+00:00"},{"alias_kind":"pith_short_8","alias_value":"HUQINWJ4","created_at":"2026-07-05T07:06:47.764583+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.01154","citing_title":"FlashDP: Private Training Large Language Models with Efficient DP-SGD","ref_index":12,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO","json":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO.json","graph_json":"https://pith.science/api/pith-number/HUQINWJ4CJ4QBRRKROF4WCZYHO/graph.json","events_json":"https://pith.science/api/pith-number/HUQINWJ4CJ4QBRRKROF4WCZYHO/events.json","paper":"https://pith.science/paper/HUQINWJ4"},"agent_actions":{"view_html":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO","download_json":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO.json","view_paper":"https://pith.science/paper/HUQINWJ4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.19215&json=true","fetch_graph":"https://pith.science/api/pith-number/HUQINWJ4CJ4QBRRKROF4WCZYHO/graph.json","fetch_events":"https://pith.science/api/pith-number/HUQINWJ4CJ4QBRRKROF4WCZYHO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO/action/storage_attestation","attest_author":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO/action/author_attestation","sign_citation":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO/action/citation_signature","submit_replication":"https://pith.science/pith/HUQINWJ4CJ4QBRRKROF4WCZYHO/action/replication_record"}},"created_at":"2026-07-05T07:06:47.764583+00:00","updated_at":"2026-07-05T07:06:47.764583+00:00"}