{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:2OMS7CNQUFOSBJWFFRML3GKQBA","short_pith_number":"pith:2OMS7CNQ","schema_version":"1.0","canonical_sha256":"d3992f89b0a15d20a6c52c58bd9950083deb23a973332db8e2e61465989b1b51","source":{"kind":"arxiv","id":"2306.01684","version":2},"attestation_state":"computed","paper":{"title":"Harnessing large-language models to generate private synthetic text","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Alexey Kurakin, Andreas Terzis, Liam MacDermed, Natalia Ponomareva, Umar Syed","submitted_at":"2023-06-02T16:59:36Z","abstract_excerpt":"Differentially private training algorithms like DP-SGD protect sensitive training data by ensuring that trained models do not reveal private information. An alternative approach, which this paper studies, is to use a sensitive dataset to generate synthetic data that is differentially private with respect to the original data, and then non-privately training a model on the synthetic data. Doing so has several advantages: synthetic data can be reused for other tasks (including for hyper parameter tuning), retained indefinitely, and shared with third parties without sacrificing privacy. However, "},"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":"2306.01684","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-06-02T16:59:36Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"a6473c91efe4c9bf696adf8d75d3d46f8d00c517220deefb8f4a1df9eab9522a","abstract_canon_sha256":"d589370869f7dbc7560336f974dc2d4e1377a410f06add39e82c2ac0e813a5ce"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:32:18.936265Z","signature_b64":"nXvZ4qELKZU25zUDHIf4zYY5rqmu7UtKf95BOcKt2te70/2qQLiHK6pA6nm641ZI4LSxZp9Apo4VO4UUeTO8CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d3992f89b0a15d20a6c52c58bd9950083deb23a973332db8e2e61465989b1b51","last_reissued_at":"2026-07-05T07:32:18.934645Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:32:18.934645Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Harnessing large-language models to generate private synthetic text","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Alexey Kurakin, Andreas Terzis, Liam MacDermed, Natalia Ponomareva, Umar Syed","submitted_at":"2023-06-02T16:59:36Z","abstract_excerpt":"Differentially private training algorithms like DP-SGD protect sensitive training data by ensuring that trained models do not reveal private information. An alternative approach, which this paper studies, is to use a sensitive dataset to generate synthetic data that is differentially private with respect to the original data, and then non-privately training a model on the synthetic data. Doing so has several advantages: synthetic data can be reused for other tasks (including for hyper parameter tuning), retained indefinitely, and shared with third parties without sacrificing privacy. However, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.01684","kind":"arxiv","version":2},"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/2306.01684/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":"2306.01684","created_at":"2026-07-05T07:32:18.935645+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.01684v2","created_at":"2026-07-05T07:32:18.935645+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.01684","created_at":"2026-07-05T07:32:18.935645+00:00"},{"alias_kind":"pith_short_12","alias_value":"2OMS7CNQUFOS","created_at":"2026-07-05T07:32:18.935645+00:00"},{"alias_kind":"pith_short_16","alias_value":"2OMS7CNQUFOSBJWF","created_at":"2026-07-05T07:32:18.935645+00:00"},{"alias_kind":"pith_short_8","alias_value":"2OMS7CNQ","created_at":"2026-07-05T07:32:18.935645+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08681","citing_title":"Asymptotic Optimality of the High-Dimensional Gaussian Mechanism and Improved Low-Dimensional Mechanisms for Differential Privacy","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01849","citing_title":"ContinuousBench: Can Differentially Private Synthetic Text Improve Capabilities?","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17432","citing_title":"DP-SelFT: Differentially Private Selective Fine-Tuning for Large Language Models","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2407.21772","citing_title":"ShieldGemma: Generative AI Content Moderation Based on Gemma","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19015","citing_title":"FedProxy: Federated Fine-Tuning of LLMs via Proxy SLMs and Heterogeneity-Aware Fusion","ref_index":121,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04461","citing_title":"DP-OPD: Differentially Private On-Policy Distillation for Language Models","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA","json":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA.json","graph_json":"https://pith.science/api/pith-number/2OMS7CNQUFOSBJWFFRML3GKQBA/graph.json","events_json":"https://pith.science/api/pith-number/2OMS7CNQUFOSBJWFFRML3GKQBA/events.json","paper":"https://pith.science/paper/2OMS7CNQ"},"agent_actions":{"view_html":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA","download_json":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA.json","view_paper":"https://pith.science/paper/2OMS7CNQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.01684&json=true","fetch_graph":"https://pith.science/api/pith-number/2OMS7CNQUFOSBJWFFRML3GKQBA/graph.json","fetch_events":"https://pith.science/api/pith-number/2OMS7CNQUFOSBJWFFRML3GKQBA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA/action/storage_attestation","attest_author":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA/action/author_attestation","sign_citation":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA/action/citation_signature","submit_replication":"https://pith.science/pith/2OMS7CNQUFOSBJWFFRML3GKQBA/action/replication_record"}},"created_at":"2026-07-05T07:32:18.935645+00:00","updated_at":"2026-07-05T07:32:18.935645+00:00"}