{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:LGJGVBH4ZBMEWCP5YEWSNFIFLQ","short_pith_number":"pith:LGJGVBH4","canonical_record":{"source":{"id":"2305.13865","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-23T09:36:58Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"14b849c063354b90f3c2dcd7e3f4068e844716775271b07d7edead27246f9449","abstract_canon_sha256":"de14f177d728a5f6b83b9150b0c53c6a46d3148da65a9a32de2a1c8c0a8063bc"},"schema_version":"1.0"},"canonical_sha256":"59926a84fcc8584b09fdc12d2695055c3698a68d3a4c75923abc9b0ed5c2969b","source":{"kind":"arxiv","id":"2305.13865","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2305.13865","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"arxiv_version","alias_value":"2305.13865v3","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.13865","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"pith_short_12","alias_value":"LGJGVBH4ZBME","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"pith_short_16","alias_value":"LGJGVBH4ZBMEWCP5","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"pith_short_8","alias_value":"LGJGVBH4","created_at":"2026-07-05T08:38:51Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:LGJGVBH4ZBMEWCP5YEWSNFIFLQ","target":"record","payload":{"canonical_record":{"source":{"id":"2305.13865","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-23T09:36:58Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"14b849c063354b90f3c2dcd7e3f4068e844716775271b07d7edead27246f9449","abstract_canon_sha256":"de14f177d728a5f6b83b9150b0c53c6a46d3148da65a9a32de2a1c8c0a8063bc"},"schema_version":"1.0"},"canonical_sha256":"59926a84fcc8584b09fdc12d2695055c3698a68d3a4c75923abc9b0ed5c2969b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:38:51.926566Z","signature_b64":"VWqA1yZoQyMhsDPJ84QB07qXtN/vRzifZ0PKrkn3+l5TOnUStk1ME4B0U/uuRqwCgZe6c/4q2R5q03doFW7gDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"59926a84fcc8584b09fdc12d2695055c3698a68d3a4c75923abc9b0ed5c2969b","last_reissued_at":"2026-07-05T08:38:51.926036Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:38:51.926036Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2305.13865","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:38:51Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YMzPRoIBL/RZbckwXoGi00cNOtE6XZtBniCECqwIeXHoGhC+bSMUXI4B65FTJdVIGQhwXUCOFsAZjxDGS2TsAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T06:16:28.543333Z"},"content_sha256":"efa4e9bc49612754edf2a569af14812afb4327bb1e2844816bf0eb75b4444ff0","schema_version":"1.0","event_id":"sha256:efa4e9bc49612754edf2a569af14812afb4327bb1e2844816bf0eb75b4444ff0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:LGJGVBH4ZBMEWCP5YEWSNFIFLQ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Selective Pre-training for Private Fine-tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Da Yu, Huishuai Zhang, Janardhan Kulkarni, Jian Yin, Saurabh Naik, Sivakanth Gopi, Tomasz Lukasz Religa, Zinan Lin","submitted_at":"2023-05-23T09:36:58Z","abstract_excerpt":"Text prediction models, when used in applications like email clients or word processors, must protect user data privacy and adhere to model size constraints. These constraints are crucial to meet memory and inference time requirements, as well as to reduce inference costs. Building small, fast, and private domain-specific language models is a thriving area of research. In this work, we show that a careful pre-training on a \\emph{subset} of the public dataset that is guided by the private dataset is crucial to train small language models with differential privacy. On standard benchmarks, small "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.13865","kind":"arxiv","version":3},"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/2305.13865/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:38:51Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"DloABU7Z9vmvYkqlFSot6nNQjwx/rehC2080L/qkz12OQXvTTLH3UCqcG1PUN5ZPt69H2+PmDWwOgGVXQvDoDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T06:16:28.543882Z"},"content_sha256":"7365fbf82fa91ee9b2d6160bdea112567d089fb5104faa1e03a9907a442cf302","schema_version":"1.0","event_id":"sha256:7365fbf82fa91ee9b2d6160bdea112567d089fb5104faa1e03a9907a442cf302"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/LGJGVBH4ZBMEWCP5YEWSNFIFLQ/bundle.json","state_url":"https://pith.science/pith/LGJGVBH4ZBMEWCP5YEWSNFIFLQ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/LGJGVBH4ZBMEWCP5YEWSNFIFLQ/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":"2026-08-05T06:16:28Z","links":{"resolver":"https://pith.science/pith/LGJGVBH4ZBMEWCP5YEWSNFIFLQ","bundle":"https://pith.science/pith/LGJGVBH4ZBMEWCP5YEWSNFIFLQ/bundle.json","state":"https://pith.science/pith/LGJGVBH4ZBMEWCP5YEWSNFIFLQ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/LGJGVBH4ZBMEWCP5YEWSNFIFLQ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:LGJGVBH4ZBMEWCP5YEWSNFIFLQ","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":"de14f177d728a5f6b83b9150b0c53c6a46d3148da65a9a32de2a1c8c0a8063bc","cross_cats_sorted":["cs.CR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-23T09:36:58Z","title_canon_sha256":"14b849c063354b90f3c2dcd7e3f4068e844716775271b07d7edead27246f9449"},"schema_version":"1.0","source":{"id":"2305.13865","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2305.13865","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"arxiv_version","alias_value":"2305.13865v3","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.13865","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"pith_short_12","alias_value":"LGJGVBH4ZBME","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"pith_short_16","alias_value":"LGJGVBH4ZBMEWCP5","created_at":"2026-07-05T08:38:51Z"},{"alias_kind":"pith_short_8","alias_value":"LGJGVBH4","created_at":"2026-07-05T08:38:51Z"}],"graph_snapshots":[{"event_id":"sha256:7365fbf82fa91ee9b2d6160bdea112567d089fb5104faa1e03a9907a442cf302","target":"graph","created_at":"2026-07-05T08:38:51Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2305.13865/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Text prediction models, when used in applications like email clients or word processors, must protect user data privacy and adhere to model size constraints. These constraints are crucial to meet memory and inference time requirements, as well as to reduce inference costs. Building small, fast, and private domain-specific language models is a thriving area of research. In this work, we show that a careful pre-training on a \\emph{subset} of the public dataset that is guided by the private dataset is crucial to train small language models with differential privacy. On standard benchmarks, small ","authors_text":"Da Yu, Huishuai Zhang, Janardhan Kulkarni, Jian Yin, Saurabh Naik, Sivakanth Gopi, Tomasz Lukasz Religa, Zinan Lin","cross_cats":["cs.CR"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-23T09:36:58Z","title":"Selective Pre-training for Private Fine-tuning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.13865","kind":"arxiv","version":3},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:efa4e9bc49612754edf2a569af14812afb4327bb1e2844816bf0eb75b4444ff0","target":"record","created_at":"2026-07-05T08:38:51Z","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":"de14f177d728a5f6b83b9150b0c53c6a46d3148da65a9a32de2a1c8c0a8063bc","cross_cats_sorted":["cs.CR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-23T09:36:58Z","title_canon_sha256":"14b849c063354b90f3c2dcd7e3f4068e844716775271b07d7edead27246f9449"},"schema_version":"1.0","source":{"id":"2305.13865","kind":"arxiv","version":3}},"canonical_sha256":"59926a84fcc8584b09fdc12d2695055c3698a68d3a4c75923abc9b0ed5c2969b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"59926a84fcc8584b09fdc12d2695055c3698a68d3a4c75923abc9b0ed5c2969b","first_computed_at":"2026-07-05T08:38:51.926036Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:38:51.926036Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"VWqA1yZoQyMhsDPJ84QB07qXtN/vRzifZ0PKrkn3+l5TOnUStk1ME4B0U/uuRqwCgZe6c/4q2R5q03doFW7gDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T08:38:51.926566Z","signed_message":"canonical_sha256_bytes"},"source_id":"2305.13865","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:efa4e9bc49612754edf2a569af14812afb4327bb1e2844816bf0eb75b4444ff0","sha256:7365fbf82fa91ee9b2d6160bdea112567d089fb5104faa1e03a9907a442cf302"],"state_sha256":"52395b1304cdeab1a6b513bc6daef3c3549d92b0f5e36b84953e065a3850ad0c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"4gWvWwRH5pjrBry6nsfM+EHTXUteaBRW2Lf32D8PTyMhHPiiuQTfmyYqXoq6Wnjr8p3A1omJfVporWxbNDK4Dg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T06:16:28.547606Z","bundle_sha256":"eaecfdfa13f6cc263bd3feee8b6067f37746bb9409680c821a24e820254f2b77"}}