{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:SOVWZQAWKXR4SFDP2T3NGF6EJG","short_pith_number":"pith:SOVWZQAW","schema_version":"1.0","canonical_sha256":"93ab6cc01655e3c9146fd4f6d317c4499b66abade00ab3960555d39168f7ce2d","source":{"kind":"arxiv","id":"2305.05973","version":3},"attestation_state":"computed","paper":{"title":"Synthetic Query Generation for Privacy-Preserving Deep Retrieval Systems using Differentially Private Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR","cs.IR"],"primary_cat":"cs.CL","authors_text":"Aldo Gael Carranza, Alex Kurakin, Matthew Jagielski, Milad Nasr, Natalia Ponomareva, Rezsa Farahani","submitted_at":"2023-05-10T08:30:31Z","abstract_excerpt":"We address the challenge of ensuring differential privacy (DP) guarantees in training deep retrieval systems. Training these systems often involves the use of contrastive-style losses, which are typically non-per-example decomposable, making them difficult to directly DP-train with since common techniques require per-example gradients. To address this issue, we propose an approach that prioritizes ensuring query privacy prior to training a deep retrieval system. Our method employs DP language models (LMs) to generate private synthetic queries representative of the original data. These syntheti"},"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":"2305.05973","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-10T08:30:31Z","cross_cats_sorted":["cs.CR","cs.IR"],"title_canon_sha256":"0da669f78f94da951d5e6682d6cbbf6a80d253677579b336bf3f47895f8d3894","abstract_canon_sha256":"cf1a9dccd8847d4f90580ee9227a8c73cb9a09b7943ff0ae9a9459b432341c4a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:21:38.783554Z","signature_b64":"MhNE0JH6g0BztnqEFsYKjQGTsa6bRMisFvkPGYFzqnw5w+F40HMtK4U8pRyg3jNTEfXfk49MEKk7v3zmTOgaAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"93ab6cc01655e3c9146fd4f6d317c4499b66abade00ab3960555d39168f7ce2d","last_reissued_at":"2026-07-05T08:21:38.783055Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:21:38.783055Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Synthetic Query Generation for Privacy-Preserving Deep Retrieval Systems using Differentially Private Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR","cs.IR"],"primary_cat":"cs.CL","authors_text":"Aldo Gael Carranza, Alex Kurakin, Matthew Jagielski, Milad Nasr, Natalia Ponomareva, Rezsa Farahani","submitted_at":"2023-05-10T08:30:31Z","abstract_excerpt":"We address the challenge of ensuring differential privacy (DP) guarantees in training deep retrieval systems. Training these systems often involves the use of contrastive-style losses, which are typically non-per-example decomposable, making them difficult to directly DP-train with since common techniques require per-example gradients. To address this issue, we propose an approach that prioritizes ensuring query privacy prior to training a deep retrieval system. Our method employs DP language models (LMs) to generate private synthetic queries representative of the original data. These syntheti"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.05973","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/2305.05973/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":"2305.05973","created_at":"2026-07-05T08:21:38.783115+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.05973v3","created_at":"2026-07-05T08:21:38.783115+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.05973","created_at":"2026-07-05T08:21:38.783115+00:00"},{"alias_kind":"pith_short_12","alias_value":"SOVWZQAWKXR4","created_at":"2026-07-05T08:21:38.783115+00:00"},{"alias_kind":"pith_short_16","alias_value":"SOVWZQAWKXR4SFDP","created_at":"2026-07-05T08:21:38.783115+00:00"},{"alias_kind":"pith_short_8","alias_value":"SOVWZQAW","created_at":"2026-07-05T08:21:38.783115+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2401.05561","citing_title":"TrustLLM: Trustworthiness in Large Language Models","ref_index":284,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG","json":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG.json","graph_json":"https://pith.science/api/pith-number/SOVWZQAWKXR4SFDP2T3NGF6EJG/graph.json","events_json":"https://pith.science/api/pith-number/SOVWZQAWKXR4SFDP2T3NGF6EJG/events.json","paper":"https://pith.science/paper/SOVWZQAW"},"agent_actions":{"view_html":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG","download_json":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG.json","view_paper":"https://pith.science/paper/SOVWZQAW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.05973&json=true","fetch_graph":"https://pith.science/api/pith-number/SOVWZQAWKXR4SFDP2T3NGF6EJG/graph.json","fetch_events":"https://pith.science/api/pith-number/SOVWZQAWKXR4SFDP2T3NGF6EJG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG/action/storage_attestation","attest_author":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG/action/author_attestation","sign_citation":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG/action/citation_signature","submit_replication":"https://pith.science/pith/SOVWZQAWKXR4SFDP2T3NGF6EJG/action/replication_record"}},"created_at":"2026-07-05T08:21:38.783115+00:00","updated_at":"2026-07-05T08:21:38.783115+00:00"}