{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:7G47O2GGJKRJJH2KM76W4ZSGW6","short_pith_number":"pith:7G47O2GG","canonical_record":{"source":{"id":"2605.22373","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-21T12:05:22Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"8bad211d4a1ec0635ef2b17c8903c168945b2838a4a569dad8479d496315a292","abstract_canon_sha256":"111d161d8a84ee44df42b578168b6fe2c010bb82c81c101878242986bb3f3013"},"schema_version":"1.0"},"canonical_sha256":"f9b9f768c64aa2949f4a67fd6e6646b781aa63f64351acd12228966bd37de6c1","source":{"kind":"arxiv","id":"2605.22373","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.22373","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"arxiv_version","alias_value":"2605.22373v1","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22373","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"pith_short_12","alias_value":"7G47O2GGJKRJ","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"pith_short_16","alias_value":"7G47O2GGJKRJJH2K","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"pith_short_8","alias_value":"7G47O2GG","created_at":"2026-05-22T01:04:40Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:7G47O2GGJKRJJH2KM76W4ZSGW6","target":"record","payload":{"canonical_record":{"source":{"id":"2605.22373","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-21T12:05:22Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"8bad211d4a1ec0635ef2b17c8903c168945b2838a4a569dad8479d496315a292","abstract_canon_sha256":"111d161d8a84ee44df42b578168b6fe2c010bb82c81c101878242986bb3f3013"},"schema_version":"1.0"},"canonical_sha256":"f9b9f768c64aa2949f4a67fd6e6646b781aa63f64351acd12228966bd37de6c1","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:04:40.466625Z","signature_b64":"yCG8nMliniusqPquhPolvxbn4qJ8lO+wIoqdcX/8Qe8qfdnqR9w28pkolrGm9GkIlOuIrkftY8nYqgJL70KoCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f9b9f768c64aa2949f4a67fd6e6646b781aa63f64351acd12228966bd37de6c1","last_reissued_at":"2026-05-22T01:04:40.465856Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:04:40.465856Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.22373","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-22T01:04:40Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JrG6OwUaNCO1IKuBAVVtav26ie+Tg82mjDKB8rno5fJ6XB7iDobtL+6MDST0NW/G7B6csjcvEtN6LDNKsB13CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-23T05:18:07.261597Z"},"content_sha256":"f293632b98164f692f580619f76b64a4620328064dd572f0acde0a817bc14e12","schema_version":"1.0","event_id":"sha256:f293632b98164f692f580619f76b64a4620328064dd572f0acde0a817bc14e12"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:7G47O2GGJKRJJH2KM76W4ZSGW6","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Boundary-targeted Membership Inference Attacks on Safety Classifiers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Adam Perer, Alexander Goldberg, Anthony Hughes, Nikolaos Aletras, Niloofar Mireshghallah, Prince Jha","submitted_at":"2026-05-21T12:05:22Z","abstract_excerpt":"Safety classifiers are essential safeguards within generative AI systems, filtering harmful content or identifying at-risk users when interacting with large language models. Despite their necessity, these models are trained on sensitive datasets including discussions of self-harm and mental health, raising important, yet poorly understood, privacy concerns. Membership inference attacks (MIAs) allow adversaries to infer membership of examples used to train models. In this work, we hypothesize that identifying the examples on which the classifier is least confident are informative for an adversa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22373","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/2605.22373/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":"2026-05-22T01:04:40Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"pZtv8pE03T8/aCgoq3GvUxNvxYZIENIEXxIKcRIYb8rChgi6Y48N2WflZsBZlAH3npO0FCELwq9LlkzjGkTMCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-23T05:18:07.262139Z"},"content_sha256":"d64ff633d3dc1beaf184b4cb19c97af791dc4a5d8816cad2d38ce47006c9140f","schema_version":"1.0","event_id":"sha256:d64ff633d3dc1beaf184b4cb19c97af791dc4a5d8816cad2d38ce47006c9140f"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/7G47O2GGJKRJJH2KM76W4ZSGW6/bundle.json","state_url":"https://pith.science/pith/7G47O2GGJKRJJH2KM76W4ZSGW6/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/7G47O2GGJKRJJH2KM76W4ZSGW6/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":"2026-05-23T05:18:07Z","links":{"resolver":"https://pith.science/pith/7G47O2GGJKRJJH2KM76W4ZSGW6","bundle":"https://pith.science/pith/7G47O2GGJKRJJH2KM76W4ZSGW6/bundle.json","state":"https://pith.science/pith/7G47O2GGJKRJJH2KM76W4ZSGW6/state.json","well_known_bundle":"https://pith.science/.well-known/pith/7G47O2GGJKRJJH2KM76W4ZSGW6/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:7G47O2GGJKRJJH2KM76W4ZSGW6","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":"111d161d8a84ee44df42b578168b6fe2c010bb82c81c101878242986bb3f3013","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-21T12:05:22Z","title_canon_sha256":"8bad211d4a1ec0635ef2b17c8903c168945b2838a4a569dad8479d496315a292"},"schema_version":"1.0","source":{"id":"2605.22373","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.22373","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"arxiv_version","alias_value":"2605.22373v1","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22373","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"pith_short_12","alias_value":"7G47O2GGJKRJ","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"pith_short_16","alias_value":"7G47O2GGJKRJJH2K","created_at":"2026-05-22T01:04:40Z"},{"alias_kind":"pith_short_8","alias_value":"7G47O2GG","created_at":"2026-05-22T01:04:40Z"}],"graph_snapshots":[{"event_id":"sha256:d64ff633d3dc1beaf184b4cb19c97af791dc4a5d8816cad2d38ce47006c9140f","target":"graph","created_at":"2026-05-22T01:04:40Z","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/2605.22373/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Safety classifiers are essential safeguards within generative AI systems, filtering harmful content or identifying at-risk users when interacting with large language models. Despite their necessity, these models are trained on sensitive datasets including discussions of self-harm and mental health, raising important, yet poorly understood, privacy concerns. Membership inference attacks (MIAs) allow adversaries to infer membership of examples used to train models. In this work, we hypothesize that identifying the examples on which the classifier is least confident are informative for an adversa","authors_text":"Adam Perer, Alexander Goldberg, Anthony Hughes, Nikolaos Aletras, Niloofar Mireshghallah, Prince Jha","cross_cats":["cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-21T12:05:22Z","title":"Boundary-targeted Membership Inference Attacks on Safety Classifiers"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22373","kind":"arxiv","version":1},"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:f293632b98164f692f580619f76b64a4620328064dd572f0acde0a817bc14e12","target":"record","created_at":"2026-05-22T01:04:40Z","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":"111d161d8a84ee44df42b578168b6fe2c010bb82c81c101878242986bb3f3013","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-21T12:05:22Z","title_canon_sha256":"8bad211d4a1ec0635ef2b17c8903c168945b2838a4a569dad8479d496315a292"},"schema_version":"1.0","source":{"id":"2605.22373","kind":"arxiv","version":1}},"canonical_sha256":"f9b9f768c64aa2949f4a67fd6e6646b781aa63f64351acd12228966bd37de6c1","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f9b9f768c64aa2949f4a67fd6e6646b781aa63f64351acd12228966bd37de6c1","first_computed_at":"2026-05-22T01:04:40.465856Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-22T01:04:40.465856Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"yCG8nMliniusqPquhPolvxbn4qJ8lO+wIoqdcX/8Qe8qfdnqR9w28pkolrGm9GkIlOuIrkftY8nYqgJL70KoCg==","signature_status":"signed_v1","signed_at":"2026-05-22T01:04:40.466625Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.22373","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f293632b98164f692f580619f76b64a4620328064dd572f0acde0a817bc14e12","sha256:d64ff633d3dc1beaf184b4cb19c97af791dc4a5d8816cad2d38ce47006c9140f"],"state_sha256":"6b65006a17a21c05a80d303a17f038aa5ad2651ca7c70842ca76c8466f22473b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"q3QvrUF9mbZ1AU76cidaCK6N1BERbySBjVr3X518HVYW2+afi636KOrC02h5CZ36P/6/sRwii/Hinl3dNG9NAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-23T05:18:07.264815Z","bundle_sha256":"b40bd534267242227d491ab54c4f2af9d05d0d1774a58c1b3139f766054f860b"}}