{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EHNUM5EZRGAOKPN5TSBKXHLYSC","short_pith_number":"pith:EHNUM5EZ","schema_version":"1.0","canonical_sha256":"21db4674998980e53dbd9c82ab9d7890a1bb70aebd5e758f3295782cda0eaf6a","source":{"kind":"arxiv","id":"2605.30808","version":1},"attestation_state":"computed","paper":{"title":"Differentially Private Preference Data Synthesis for Large Language Model Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Fengyu Gao, Jing Yang","submitted_at":"2026-05-29T03:53:12Z","abstract_excerpt":"Preference alignment is a crucial post-training step for large language models (LLMs) to ensure their outputs align with human values. However, post-training on real human preference data raises privacy concerns, as these datasets often contain sensitive user prompts and human judgments. To address this, we propose DPPrefSyn, a novel algorithm for generating differentially private (DP) synthetic preference data to enable privacy-preserving preference alignment. DPPrefSyn is a principled framework grounded in the Bradley-Terry preference model and the intrinsic geometric structure of pairwise 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":"2605.30808","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-05-29T03:53:12Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"2c35dcdddeb849bc34ae6a67567a5ceda4f944da2e0efb726edee9e23f34c258","abstract_canon_sha256":"78a8377e333df1edb9eabf274bea8a47aff5c2a89082613ed7a72088f969df5e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:03:17.979019Z","signature_b64":"zp6lAc9iH0jhEb5MtIZR/66VwDjLskSmFzhdXqXto61hDZO4F7eTgsVoJtw4GL0QCXEnUjO8334E+4ZBWOWAAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"21db4674998980e53dbd9c82ab9d7890a1bb70aebd5e758f3295782cda0eaf6a","last_reissued_at":"2026-06-01T01:03:17.978040Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:03:17.978040Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Differentially Private Preference Data Synthesis for Large Language Model Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Fengyu Gao, Jing Yang","submitted_at":"2026-05-29T03:53:12Z","abstract_excerpt":"Preference alignment is a crucial post-training step for large language models (LLMs) to ensure their outputs align with human values. However, post-training on real human preference data raises privacy concerns, as these datasets often contain sensitive user prompts and human judgments. To address this, we propose DPPrefSyn, a novel algorithm for generating differentially private (DP) synthetic preference data to enable privacy-preserving preference alignment. DPPrefSyn is a principled framework grounded in the Bradley-Terry preference model and the intrinsic geometric structure of pairwise h"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.30808","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/2605.30808/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":"2605.30808","created_at":"2026-06-01T01:03:17.978205+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.30808v1","created_at":"2026-06-01T01:03:17.978205+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.30808","created_at":"2026-06-01T01:03:17.978205+00:00"},{"alias_kind":"pith_short_12","alias_value":"EHNUM5EZRGAO","created_at":"2026-06-01T01:03:17.978205+00:00"},{"alias_kind":"pith_short_16","alias_value":"EHNUM5EZRGAOKPN5","created_at":"2026-06-01T01:03:17.978205+00:00"},{"alias_kind":"pith_short_8","alias_value":"EHNUM5EZ","created_at":"2026-06-01T01:03:17.978205+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC","json":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC.json","graph_json":"https://pith.science/api/pith-number/EHNUM5EZRGAOKPN5TSBKXHLYSC/graph.json","events_json":"https://pith.science/api/pith-number/EHNUM5EZRGAOKPN5TSBKXHLYSC/events.json","paper":"https://pith.science/paper/EHNUM5EZ"},"agent_actions":{"view_html":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC","download_json":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC.json","view_paper":"https://pith.science/paper/EHNUM5EZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.30808&json=true","fetch_graph":"https://pith.science/api/pith-number/EHNUM5EZRGAOKPN5TSBKXHLYSC/graph.json","fetch_events":"https://pith.science/api/pith-number/EHNUM5EZRGAOKPN5TSBKXHLYSC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC/action/storage_attestation","attest_author":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC/action/author_attestation","sign_citation":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC/action/citation_signature","submit_replication":"https://pith.science/pith/EHNUM5EZRGAOKPN5TSBKXHLYSC/action/replication_record"}},"created_at":"2026-06-01T01:03:17.978205+00:00","updated_at":"2026-06-01T01:03:17.978205+00:00"}