{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:7EISJFCL2345OCUMBOCB4R2EYA","short_pith_number":"pith:7EISJFCL","schema_version":"1.0","canonical_sha256":"f91124944bd6f9d70a8c0b841e4744c00db9e79fb23bf15fce4f73f3da3b290e","source":{"kind":"arxiv","id":"2204.07667","version":3},"attestation_state":"computed","paper":{"title":"Just Fine-tune Twice: Selective Differential Privacy for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.CL","authors_text":"Chiyuan Zhang, Ruoxi Jia, Ryan Shea, Si Chen, Weiyan Shi, Zhou Yu","submitted_at":"2022-04-15T22:36:55Z","abstract_excerpt":"Protecting large language models from privacy leakage is becoming increasingly crucial with their wide adoption in real-world products. Yet applying differential privacy (DP), a canonical notion with provable privacy guarantees for machine learning models, to those models remains challenging due to the trade-off between model utility and privacy loss. Utilizing the fact that sensitive information in language data tends to be sparse, Shi et al. (2021) formalized a DP notion extension called Selective Differential Privacy (SDP) to protect only the sensitive tokens defined by a policy function. H"},"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":"2204.07667","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-04-15T22:36:55Z","cross_cats_sorted":["cs.AI","cs.CR"],"title_canon_sha256":"e458eaae105739441f98551ca798d8ca77e9d075ae02db50bef71c24a60c2e50","abstract_canon_sha256":"aacbd54b301b357c3f5a74dcce2b1081462e7d417af0823e33136a137d528296"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:10:55.720286Z","signature_b64":"J/IFjLWtw2n6Ni5S50zKyucp35TLtQZyGiHTtoIzDnBjWGoCmX+Xd3DbNSS3eg0cgZz62LX1h6IDBJFBRYegAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f91124944bd6f9d70a8c0b841e4744c00db9e79fb23bf15fce4f73f3da3b290e","last_reissued_at":"2026-07-05T05:10:55.719852Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:10:55.719852Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Just Fine-tune Twice: Selective Differential Privacy for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.CL","authors_text":"Chiyuan Zhang, Ruoxi Jia, Ryan Shea, Si Chen, Weiyan Shi, Zhou Yu","submitted_at":"2022-04-15T22:36:55Z","abstract_excerpt":"Protecting large language models from privacy leakage is becoming increasingly crucial with their wide adoption in real-world products. Yet applying differential privacy (DP), a canonical notion with provable privacy guarantees for machine learning models, to those models remains challenging due to the trade-off between model utility and privacy loss. Utilizing the fact that sensitive information in language data tends to be sparse, Shi et al. (2021) formalized a DP notion extension called Selective Differential Privacy (SDP) to protect only the sensitive tokens defined by a policy function. H"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2204.07667","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/2204.07667/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":"2204.07667","created_at":"2026-07-05T05:10:55.719906+00:00"},{"alias_kind":"arxiv_version","alias_value":"2204.07667v3","created_at":"2026-07-05T05:10:55.719906+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2204.07667","created_at":"2026-07-05T05:10:55.719906+00:00"},{"alias_kind":"pith_short_12","alias_value":"7EISJFCL2345","created_at":"2026-07-05T05:10:55.719906+00:00"},{"alias_kind":"pith_short_16","alias_value":"7EISJFCL2345OCUM","created_at":"2026-07-05T05:10:55.719906+00:00"},{"alias_kind":"pith_short_8","alias_value":"7EISJFCL","created_at":"2026-07-05T05:10:55.719906+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.23883","citing_title":"Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges","ref_index":140,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA","json":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA.json","graph_json":"https://pith.science/api/pith-number/7EISJFCL2345OCUMBOCB4R2EYA/graph.json","events_json":"https://pith.science/api/pith-number/7EISJFCL2345OCUMBOCB4R2EYA/events.json","paper":"https://pith.science/paper/7EISJFCL"},"agent_actions":{"view_html":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA","download_json":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA.json","view_paper":"https://pith.science/paper/7EISJFCL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2204.07667&json=true","fetch_graph":"https://pith.science/api/pith-number/7EISJFCL2345OCUMBOCB4R2EYA/graph.json","fetch_events":"https://pith.science/api/pith-number/7EISJFCL2345OCUMBOCB4R2EYA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA/action/storage_attestation","attest_author":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA/action/author_attestation","sign_citation":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA/action/citation_signature","submit_replication":"https://pith.science/pith/7EISJFCL2345OCUMBOCB4R2EYA/action/replication_record"}},"created_at":"2026-07-05T05:10:55.719906+00:00","updated_at":"2026-07-05T05:10:55.719906+00:00"}