{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:CCFHPOTSAETUW4PC7X53PFKPQP","short_pith_number":"pith:CCFHPOTS","schema_version":"1.0","canonical_sha256":"108a77ba7201274b71e2fdfbb7954f83c046298f7b4cf02d1baafbf43546411e","source":{"kind":"arxiv","id":"2501.18532","version":2},"attestation_state":"computed","paper":{"title":"Differentially Private Steering for Large Language Model Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Amartya Sanyal, Anmol Goel, Iryna Gurevych, Yaxi Hu","submitted_at":"2025-01-30T17:58:36Z","abstract_excerpt":"Aligning Large Language Models (LLMs) with human values and away from undesirable behaviors (such as hallucination) has become increasingly important. Recently, steering LLMs towards a desired behavior via activation editing has emerged as an effective method to mitigate harmful generations at inference-time. Activation editing modifies LLM representations by preserving information from positive demonstrations (e.g., truthful) and minimising information from negative demonstrations (e.g., hallucinations). When these demonstrations come from a private dataset, the aligned LLM may leak private i"},"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":"2501.18532","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-30T17:58:36Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"d023869629fcf99ad66adffb275baff89c03fa5d15d6d3ceba820d08c8407df9","abstract_canon_sha256":"d67f737a0814f287f80ed66d38622efa3faf1737c10a100dfb2acd4e3000e9dd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:35:45.029313Z","signature_b64":"bRzYGdc4V2tMA/4qvDRYserkjZ2ErVth9Ud05LjmslzvTyV9kjSfZIY8kUSU4s21ZxqCd+KyXUzJr7K23KTICg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"108a77ba7201274b71e2fdfbb7954f83c046298f7b4cf02d1baafbf43546411e","last_reissued_at":"2026-07-05T10:35:45.028722Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:35:45.028722Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Differentially Private Steering for Large Language Model Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Amartya Sanyal, Anmol Goel, Iryna Gurevych, Yaxi Hu","submitted_at":"2025-01-30T17:58:36Z","abstract_excerpt":"Aligning Large Language Models (LLMs) with human values and away from undesirable behaviors (such as hallucination) has become increasingly important. Recently, steering LLMs towards a desired behavior via activation editing has emerged as an effective method to mitigate harmful generations at inference-time. Activation editing modifies LLM representations by preserving information from positive demonstrations (e.g., truthful) and minimising information from negative demonstrations (e.g., hallucinations). When these demonstrations come from a private dataset, the aligned LLM may leak private i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.18532","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/2501.18532/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":"2501.18532","created_at":"2026-07-05T10:35:45.028799+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.18532v2","created_at":"2026-07-05T10:35:45.028799+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.18532","created_at":"2026-07-05T10:35:45.028799+00:00"},{"alias_kind":"pith_short_12","alias_value":"CCFHPOTSAETU","created_at":"2026-07-05T10:35:45.028799+00:00"},{"alias_kind":"pith_short_16","alias_value":"CCFHPOTSAETUW4PC","created_at":"2026-07-05T10:35:45.028799+00:00"},{"alias_kind":"pith_short_8","alias_value":"CCFHPOTS","created_at":"2026-07-05T10:35:45.028799+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/CCFHPOTSAETUW4PC7X53PFKPQP","json":"https://pith.science/pith/CCFHPOTSAETUW4PC7X53PFKPQP.json","graph_json":"https://pith.science/api/pith-number/CCFHPOTSAETUW4PC7X53PFKPQP/graph.json","events_json":"https://pith.science/api/pith-number/CCFHPOTSAETUW4PC7X53PFKPQP/events.json","paper":"https://pith.science/paper/CCFHPOTS"},"agent_actions":{"view_html":"https://pith.science/pith/CCFHPOTSAETUW4PC7X53PFKPQP","download_json":"https://pith.science/pith/CCFHPOTSAETUW4PC7X53PFKPQP.json","view_paper":"https://pith.science/paper/CCFHPOTS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.18532&json=true","fetch_graph":"https://pith.science/api/pith-number/CCFHPOTSAETUW4PC7X53PFKPQP/graph.json","fetch_events":"https://pith.science/api/pith-number/CCFHPOTSAETUW4PC7X53PFKPQP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CCFHPOTSAETUW4PC7X53PFKPQP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CCFHPOTSAETUW4PC7X53PFKPQP/action/storage_attestation","attest_author":"https://pith.science/pith/CCFHPOTSAETUW4PC7X53PFKPQP/action/author_attestation","sign_citation":"https://pith.science/pith/CCFHPOTSAETUW4PC7X53PFKPQP/action/citation_signature","submit_replication":"https://pith.science/pith/CCFHPOTSAETUW4PC7X53PFKPQP/action/replication_record"}},"created_at":"2026-07-05T10:35:45.028799+00:00","updated_at":"2026-07-05T10:35:45.028799+00:00"}