{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:OZJVTBWZDQY2RGVV65HCUJY5EL","short_pith_number":"pith:OZJVTBWZ","schema_version":"1.0","canonical_sha256":"76535986d91c31a89ab5f74e2a271d22df7053aaad443d9a54d17baac4aa3d54","source":{"kind":"arxiv","id":"2606.10481","version":1},"attestation_state":"computed","paper":{"title":"Advancing the State-of-the-Art in Empirical Privacy Auditing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Arun Ganesh, Brendan McMahan, Galen Andrew, Nicole Mitchell, Peter Kairouz","submitted_at":"2026-06-09T06:50:49Z","abstract_excerpt":"Parameter-efficient fine-tuning of large language models (LLMs) can exhibit problematic memorization of individual training examples. Empirical privacy auditing (EPA) quantifies this risk by measuring realistic data leakage on membership inference (MI) or reconstruction attacks. A key challenge in EPA is designing ``canary'' examples that are mixed with the privacy-sensitive training data. We propose generating synthetic canaries via high-temperature sampling ($T \\geq 0.8$) from LLMs, using prompts tailored to the privacy-sensitive training data. These canaries act as high-influence outliers, "},"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":"2606.10481","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-09T06:50:49Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CR","stat.ML"],"title_canon_sha256":"4e1c45581b4c1a698c93a20ca70737ab7ca9fd9087482d0cdc1c8f597ed5213f","abstract_canon_sha256":"09d8d660815e526cf0c367bdb044c543874343029c630d8be633c4b76331685a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-10T01:10:21.503623Z","signature_b64":"ps2Q+Zc7m3isoJ22MejueHXqNK2/RN11sK0j9q81U9u5UuaAaP48WJdFdz8YUXRnf1V48cNj2QyrMFqd/LDSDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"76535986d91c31a89ab5f74e2a271d22df7053aaad443d9a54d17baac4aa3d54","last_reissued_at":"2026-06-10T01:10:21.502854Z","signature_status":"signed_v1","first_computed_at":"2026-06-10T01:10:21.502854Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Advancing the State-of-the-Art in Empirical Privacy Auditing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Arun Ganesh, Brendan McMahan, Galen Andrew, Nicole Mitchell, Peter Kairouz","submitted_at":"2026-06-09T06:50:49Z","abstract_excerpt":"Parameter-efficient fine-tuning of large language models (LLMs) can exhibit problematic memorization of individual training examples. Empirical privacy auditing (EPA) quantifies this risk by measuring realistic data leakage on membership inference (MI) or reconstruction attacks. A key challenge in EPA is designing ``canary'' examples that are mixed with the privacy-sensitive training data. We propose generating synthetic canaries via high-temperature sampling ($T \\geq 0.8$) from LLMs, using prompts tailored to the privacy-sensitive training data. These canaries act as high-influence outliers, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.10481","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/2606.10481/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":"2606.10481","created_at":"2026-06-10T01:10:21.502988+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.10481v1","created_at":"2026-06-10T01:10:21.502988+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.10481","created_at":"2026-06-10T01:10:21.502988+00:00"},{"alias_kind":"pith_short_12","alias_value":"OZJVTBWZDQY2","created_at":"2026-06-10T01:10:21.502988+00:00"},{"alias_kind":"pith_short_16","alias_value":"OZJVTBWZDQY2RGVV","created_at":"2026-06-10T01:10:21.502988+00:00"},{"alias_kind":"pith_short_8","alias_value":"OZJVTBWZ","created_at":"2026-06-10T01:10:21.502988+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/OZJVTBWZDQY2RGVV65HCUJY5EL","json":"https://pith.science/pith/OZJVTBWZDQY2RGVV65HCUJY5EL.json","graph_json":"https://pith.science/api/pith-number/OZJVTBWZDQY2RGVV65HCUJY5EL/graph.json","events_json":"https://pith.science/api/pith-number/OZJVTBWZDQY2RGVV65HCUJY5EL/events.json","paper":"https://pith.science/paper/OZJVTBWZ"},"agent_actions":{"view_html":"https://pith.science/pith/OZJVTBWZDQY2RGVV65HCUJY5EL","download_json":"https://pith.science/pith/OZJVTBWZDQY2RGVV65HCUJY5EL.json","view_paper":"https://pith.science/paper/OZJVTBWZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.10481&json=true","fetch_graph":"https://pith.science/api/pith-number/OZJVTBWZDQY2RGVV65HCUJY5EL/graph.json","fetch_events":"https://pith.science/api/pith-number/OZJVTBWZDQY2RGVV65HCUJY5EL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OZJVTBWZDQY2RGVV65HCUJY5EL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OZJVTBWZDQY2RGVV65HCUJY5EL/action/storage_attestation","attest_author":"https://pith.science/pith/OZJVTBWZDQY2RGVV65HCUJY5EL/action/author_attestation","sign_citation":"https://pith.science/pith/OZJVTBWZDQY2RGVV65HCUJY5EL/action/citation_signature","submit_replication":"https://pith.science/pith/OZJVTBWZDQY2RGVV65HCUJY5EL/action/replication_record"}},"created_at":"2026-06-10T01:10:21.502988+00:00","updated_at":"2026-06-10T01:10:21.502988+00:00"}