{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:ZMUTFYO7PUWXPEJ27NQX4KZZGK","short_pith_number":"pith:ZMUTFYO7","canonical_record":{"source":{"id":"2604.20985","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-22T18:13:37Z","cross_cats_sorted":["cs.AI","cs.CR","stat.ML"],"title_canon_sha256":"74cd4d4751e8f45791432b828fb58bc3d64057097560619189ad21ff1dfc8947","abstract_canon_sha256":"c8d3a2f58553318ba692d7f377561c860b961f49b658fda3d658dcd3ee6ca237"},"schema_version":"1.0"},"canonical_sha256":"cb2932e1df7d2d77913afb617e2b3932af8e1d74f61fe1db1c2fc95ac531ea6a","source":{"kind":"arxiv","id":"2604.20985","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.20985","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"arxiv_version","alias_value":"2604.20985v2","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.20985","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"pith_short_12","alias_value":"ZMUTFYO7PUWX","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"pith_short_16","alias_value":"ZMUTFYO7PUWXPEJ2","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"pith_short_8","alias_value":"ZMUTFYO7","created_at":"2026-05-21T01:04:25Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:ZMUTFYO7PUWXPEJ27NQX4KZZGK","target":"record","payload":{"canonical_record":{"source":{"id":"2604.20985","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-22T18:13:37Z","cross_cats_sorted":["cs.AI","cs.CR","stat.ML"],"title_canon_sha256":"74cd4d4751e8f45791432b828fb58bc3d64057097560619189ad21ff1dfc8947","abstract_canon_sha256":"c8d3a2f58553318ba692d7f377561c860b961f49b658fda3d658dcd3ee6ca237"},"schema_version":"1.0"},"canonical_sha256":"cb2932e1df7d2d77913afb617e2b3932af8e1d74f61fe1db1c2fc95ac531ea6a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T01:04:25.964152Z","signature_b64":"dnepvgyChzGNLVbeeYohjZrQ1tBvytbGrxlAgh73UblJxk6RkQ3fD8sNLdmnIjk72NHVUzFr+e6CjkWBk98ABQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cb2932e1df7d2d77913afb617e2b3932af8e1d74f61fe1db1c2fc95ac531ea6a","last_reissued_at":"2026-05-21T01:04:25.963335Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T01:04:25.963335Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.20985","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-21T01:04:25Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kr5+wEtsoqdcPKCzHqlPQOxXuHcKNlj1jvRRzQlfJuxb+1ELaDH6m6kppsxG6g7SrHRCOZMJMlhxE9iEAFUnCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-29T18:59:39.086967Z"},"content_sha256":"f19b8ec5c4317dd4cc57bccd660d2b06f798a7ad389c9b812d556fb226f22f6b","schema_version":"1.0","event_id":"sha256:f19b8ec5c4317dd4cc57bccd660d2b06f798a7ad389c9b812d556fb226f22f6b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:ZMUTFYO7PUWXPEJ27NQX4KZZGK","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Differentially Private Model Merging","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Post-processing existing differentially private models with random selection or linear combination produces a model meeting any target privacy level without additional training.","cross_cats":["cs.AI","cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Manzil Zaheer, Qichuan Yin, Tian Li","submitted_at":"2026-04-22T18:13:37Z","abstract_excerpt":"In machine learning, privacy requirements at inference or deployment time often evolve due to changing policies, regulations, or user preferences. In this work, we aim to construct a magnitude of models to satisfy any target differential privacy (DP) requirement without additional training, given a set of existing models trained on the same dataset with different privacy/utility tradeoffs. We propose two post-processing techniques, namely random selection and linear combination, to generate final private models satisfying any target privacy parameter. We provide privacy accounting of these app"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We propose two post processing techniques, namely random selection and linear combination, to output a final private model for any target privacy parameter. We provide privacy accounting of these approaches from the lens of Rényi DP and privacy loss distributions for general problems. In a case study on private mean estimation, we fully characterize the privacy/utility results and theoretically establish the superiority of linear combination over random selection.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The existing models are trained on the same dataset with different privacy/utility tradeoffs, which is required for the post-processing techniques to correctly compose privacy guarantees without additional training.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Post-processing via random selection or linear combination generates differentially private models for arbitrary privacy parameters from pre-trained models on the same dataset.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Post-processing existing differentially private models with random selection or linear combination produces a model meeting any target privacy level without additional training.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f4dd4840f1b51e0df8bb48daed3b13592d99603956ffab65f9375a43a5438bac"},"source":{"id":"2604.20985","kind":"arxiv","version":2},"verdict":{"id":"4ed4ce60-10a2-4ff3-a88e-1782a20a0e27","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T01:10:56.130853Z","strongest_claim":"We propose two post processing techniques, namely random selection and linear combination, to output a final private model for any target privacy parameter. We provide privacy accounting of these approaches from the lens of Rényi DP and privacy loss distributions for general problems. In a case study on private mean estimation, we fully characterize the privacy/utility results and theoretically establish the superiority of linear combination over random selection.","one_line_summary":"Post-processing via random selection or linear combination generates differentially private models for arbitrary privacy parameters from pre-trained models on the same dataset.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The existing models are trained on the same dataset with different privacy/utility tradeoffs, which is required for the post-processing techniques to correctly compose privacy guarantees without additional training.","pith_extraction_headline":"Post-processing existing differentially private models with random selection or linear combination produces a model meeting any target privacy level without additional training."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.20985/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-20T01:29:53.754074Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"482d38c29369652c0ab2d1fc2260afb03061e5dbcdd5312859ad9fa204dff060"},"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"},"verdict_id":"4ed4ce60-10a2-4ff3-a88e-1782a20a0e27"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-21T01:04:25Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"IHgsE8lJDE1lVxKsUie22oatI9bq5wXHOQhIFdfnBZc7higt+TY/bMFteFQmSj5On+v/LtYKFQ64O9M56m0eBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-29T18:59:39.087527Z"},"content_sha256":"bd1254eef06712f5bfbd3e74e49f9a002e579866715512db5e15b3eaf14a7f20","schema_version":"1.0","event_id":"sha256:bd1254eef06712f5bfbd3e74e49f9a002e579866715512db5e15b3eaf14a7f20"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/ZMUTFYO7PUWXPEJ27NQX4KZZGK/bundle.json","state_url":"https://pith.science/pith/ZMUTFYO7PUWXPEJ27NQX4KZZGK/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/ZMUTFYO7PUWXPEJ27NQX4KZZGK/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-29T18:59:39Z","links":{"resolver":"https://pith.science/pith/ZMUTFYO7PUWXPEJ27NQX4KZZGK","bundle":"https://pith.science/pith/ZMUTFYO7PUWXPEJ27NQX4KZZGK/bundle.json","state":"https://pith.science/pith/ZMUTFYO7PUWXPEJ27NQX4KZZGK/state.json","well_known_bundle":"https://pith.science/.well-known/pith/ZMUTFYO7PUWXPEJ27NQX4KZZGK/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:ZMUTFYO7PUWXPEJ27NQX4KZZGK","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"c8d3a2f58553318ba692d7f377561c860b961f49b658fda3d658dcd3ee6ca237","cross_cats_sorted":["cs.AI","cs.CR","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-22T18:13:37Z","title_canon_sha256":"74cd4d4751e8f45791432b828fb58bc3d64057097560619189ad21ff1dfc8947"},"schema_version":"1.0","source":{"id":"2604.20985","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.20985","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"arxiv_version","alias_value":"2604.20985v2","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.20985","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"pith_short_12","alias_value":"ZMUTFYO7PUWX","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"pith_short_16","alias_value":"ZMUTFYO7PUWXPEJ2","created_at":"2026-05-21T01:04:25Z"},{"alias_kind":"pith_short_8","alias_value":"ZMUTFYO7","created_at":"2026-05-21T01:04:25Z"}],"graph_snapshots":[{"event_id":"sha256:bd1254eef06712f5bfbd3e74e49f9a002e579866715512db5e15b3eaf14a7f20","target":"graph","created_at":"2026-05-21T01:04:25Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"We propose two post processing techniques, namely random selection and linear combination, to output a final private model for any target privacy parameter. We provide privacy accounting of these approaches from the lens of Rényi DP and privacy loss distributions for general problems. In a case study on private mean estimation, we fully characterize the privacy/utility results and theoretically establish the superiority of linear combination over random selection."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The existing models are trained on the same dataset with different privacy/utility tradeoffs, which is required for the post-processing techniques to correctly compose privacy guarantees without additional training."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Post-processing via random selection or linear combination generates differentially private models for arbitrary privacy parameters from pre-trained models on the same dataset."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Post-processing existing differentially private models with random selection or linear combination produces a model meeting any target privacy level without additional training."}],"snapshot_sha256":"f4dd4840f1b51e0df8bb48daed3b13592d99603956ffab65f9375a43a5438bac"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-20T01:29:53.754074Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.20985/integrity.json","findings":[],"snapshot_sha256":"482d38c29369652c0ab2d1fc2260afb03061e5dbcdd5312859ad9fa204dff060","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"In machine learning, privacy requirements at inference or deployment time often evolve due to changing policies, regulations, or user preferences. In this work, we aim to construct a magnitude of models to satisfy any target differential privacy (DP) requirement without additional training, given a set of existing models trained on the same dataset with different privacy/utility tradeoffs. We propose two post-processing techniques, namely random selection and linear combination, to generate final private models satisfying any target privacy parameter. We provide privacy accounting of these app","authors_text":"Manzil Zaheer, Qichuan Yin, Tian Li","cross_cats":["cs.AI","cs.CR","stat.ML"],"headline":"Post-processing existing differentially private models with random selection or linear combination produces a model meeting any target privacy level without additional training.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-22T18:13:37Z","title":"Differentially Private Model Merging"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.20985","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T01:10:56.130853Z","id":"4ed4ce60-10a2-4ff3-a88e-1782a20a0e27","model_set":{"reader":"grok-4.3"},"one_line_summary":"Post-processing via random selection or linear combination generates differentially private models for arbitrary privacy parameters from pre-trained models on the same dataset.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Post-processing existing differentially private models with random selection or linear combination produces a model meeting any target privacy level without additional training.","strongest_claim":"We propose two post processing techniques, namely random selection and linear combination, to output a final private model for any target privacy parameter. We provide privacy accounting of these approaches from the lens of Rényi DP and privacy loss distributions for general problems. In a case study on private mean estimation, we fully characterize the privacy/utility results and theoretically establish the superiority of linear combination over random selection.","weakest_assumption":"The existing models are trained on the same dataset with different privacy/utility tradeoffs, which is required for the post-processing techniques to correctly compose privacy guarantees without additional training."}},"verdict_id":"4ed4ce60-10a2-4ff3-a88e-1782a20a0e27"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f19b8ec5c4317dd4cc57bccd660d2b06f798a7ad389c9b812d556fb226f22f6b","target":"record","created_at":"2026-05-21T01:04:25Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"c8d3a2f58553318ba692d7f377561c860b961f49b658fda3d658dcd3ee6ca237","cross_cats_sorted":["cs.AI","cs.CR","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-22T18:13:37Z","title_canon_sha256":"74cd4d4751e8f45791432b828fb58bc3d64057097560619189ad21ff1dfc8947"},"schema_version":"1.0","source":{"id":"2604.20985","kind":"arxiv","version":2}},"canonical_sha256":"cb2932e1df7d2d77913afb617e2b3932af8e1d74f61fe1db1c2fc95ac531ea6a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"cb2932e1df7d2d77913afb617e2b3932af8e1d74f61fe1db1c2fc95ac531ea6a","first_computed_at":"2026-05-21T01:04:25.963335Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-21T01:04:25.963335Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"dnepvgyChzGNLVbeeYohjZrQ1tBvytbGrxlAgh73UblJxk6RkQ3fD8sNLdmnIjk72NHVUzFr+e6CjkWBk98ABQ==","signature_status":"signed_v1","signed_at":"2026-05-21T01:04:25.964152Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.20985","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f19b8ec5c4317dd4cc57bccd660d2b06f798a7ad389c9b812d556fb226f22f6b","sha256:bd1254eef06712f5bfbd3e74e49f9a002e579866715512db5e15b3eaf14a7f20"],"state_sha256":"c8c0fbc203085c300a5b032726f8b0e952089c542892bd0715cef00ee78072f9"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7j5Sc/y6ge4XJsQcDrGsoPxKarF/mFyMWa0Q+2pSu6UcN60Nh/PAjbF3Dwr6qlQSL8dLM4hSADhdYGqvGmHSDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-29T18:59:39.090669Z","bundle_sha256":"f9ee2292aafcbd46679273f7ccbebec72df8fd63b39456015177193e2724d1c8"}}