{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:C4Z35HQZ4DEAI47ZP5BEXCDYWN","short_pith_number":"pith:C4Z35HQZ","schema_version":"1.0","canonical_sha256":"1733be9e19e0c80473f97f424b8878b34046b766cf991983b248560c74b572cc","source":{"kind":"arxiv","id":"2505.08849","version":1},"attestation_state":"computed","paper":{"title":"Improved Algorithms for Differentially Private Language Model Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Hao Tang, Keyu Chen, Qinglin Liu, Yizhao Xu","submitted_at":"2025-05-13T16:18:59Z","abstract_excerpt":"Language model alignment is crucial for ensuring that large language models (LLMs) align with human preferences, yet it often involves sensitive user data, raising significant privacy concerns. While prior work has integrated differential privacy (DP) with alignment techniques, their performance remains limited. In this paper, we propose novel algorithms for privacy-preserving alignment and rigorously analyze their effectiveness across varying privacy budgets and models. Our framework can be deployed on two celebrated alignment techniques, namely direct preference optimization (DPO) and reinfo"},"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":"2505.08849","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2025-05-13T16:18:59Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"69685c309359d5a079e0ea2007271dd59e98cf8a2971351d911daceffab57ae1","abstract_canon_sha256":"326e60cac19acc8bfe2a3fe80d960942ff2adb75ad8bbb4acad803b6d1892245"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:02:45.451178Z","signature_b64":"rjk4X8475J6C6Y6qILaiHvc5HbYgLMampe6qi+yOZTeDbkz9K4cbB0g2DUbQMocgbg9QS+3gwq858okm17f2Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1733be9e19e0c80473f97f424b8878b34046b766cf991983b248560c74b572cc","last_reissued_at":"2026-07-05T11:02:45.450688Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:02:45.450688Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improved Algorithms for Differentially Private Language Model Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Hao Tang, Keyu Chen, Qinglin Liu, Yizhao Xu","submitted_at":"2025-05-13T16:18:59Z","abstract_excerpt":"Language model alignment is crucial for ensuring that large language models (LLMs) align with human preferences, yet it often involves sensitive user data, raising significant privacy concerns. While prior work has integrated differential privacy (DP) with alignment techniques, their performance remains limited. In this paper, we propose novel algorithms for privacy-preserving alignment and rigorously analyze their effectiveness across varying privacy budgets and models. Our framework can be deployed on two celebrated alignment techniques, namely direct preference optimization (DPO) and reinfo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.08849","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/2505.08849/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":"2505.08849","created_at":"2026-07-05T11:02:45.450746+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.08849v1","created_at":"2026-07-05T11:02:45.450746+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.08849","created_at":"2026-07-05T11:02:45.450746+00:00"},{"alias_kind":"pith_short_12","alias_value":"C4Z35HQZ4DEA","created_at":"2026-07-05T11:02:45.450746+00:00"},{"alias_kind":"pith_short_16","alias_value":"C4Z35HQZ4DEAI47Z","created_at":"2026-07-05T11:02:45.450746+00:00"},{"alias_kind":"pith_short_8","alias_value":"C4Z35HQZ","created_at":"2026-07-05T11:02:45.450746+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.11134","citing_title":"Spurious Correlation Learning in Preference Optimization: Mechanisms, Consequences, and Mitigation via Tie Training","ref_index":46,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN","json":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN.json","graph_json":"https://pith.science/api/pith-number/C4Z35HQZ4DEAI47ZP5BEXCDYWN/graph.json","events_json":"https://pith.science/api/pith-number/C4Z35HQZ4DEAI47ZP5BEXCDYWN/events.json","paper":"https://pith.science/paper/C4Z35HQZ"},"agent_actions":{"view_html":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN","download_json":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN.json","view_paper":"https://pith.science/paper/C4Z35HQZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.08849&json=true","fetch_graph":"https://pith.science/api/pith-number/C4Z35HQZ4DEAI47ZP5BEXCDYWN/graph.json","fetch_events":"https://pith.science/api/pith-number/C4Z35HQZ4DEAI47ZP5BEXCDYWN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN/action/storage_attestation","attest_author":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN/action/author_attestation","sign_citation":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN/action/citation_signature","submit_replication":"https://pith.science/pith/C4Z35HQZ4DEAI47ZP5BEXCDYWN/action/replication_record"}},"created_at":"2026-07-05T11:02:45.450746+00:00","updated_at":"2026-07-05T11:02:45.450746+00:00"}