{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2022:BJ7NM3YRPJQPC55HEX7462J3WZ","short_pith_number":"pith:BJ7NM3YR","canonical_record":{"source":{"id":"2202.06539","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2022-02-14T08:20:15Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"4e6d30a501403d539b164a84db0b93856b64a233ab359a29d18dce646972e21e","abstract_canon_sha256":"191f795825bd78e9a5fc70a749428eed511293417d27416a73301b531822d273"},"schema_version":"1.0"},"canonical_sha256":"0a7ed66f117a60f177a725ffcf693bb677b77301534a1885e42e0758b103b570","source":{"kind":"arxiv","id":"2202.06539","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2202.06539","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"arxiv_version","alias_value":"2202.06539v3","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2202.06539","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"pith_short_12","alias_value":"BJ7NM3YRPJQP","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"pith_short_16","alias_value":"BJ7NM3YRPJQPC55H","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"pith_short_8","alias_value":"BJ7NM3YR","created_at":"2026-07-05T05:26:38Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2022:BJ7NM3YRPJQPC55HEX7462J3WZ","target":"record","payload":{"canonical_record":{"source":{"id":"2202.06539","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2022-02-14T08:20:15Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"4e6d30a501403d539b164a84db0b93856b64a233ab359a29d18dce646972e21e","abstract_canon_sha256":"191f795825bd78e9a5fc70a749428eed511293417d27416a73301b531822d273"},"schema_version":"1.0"},"canonical_sha256":"0a7ed66f117a60f177a725ffcf693bb677b77301534a1885e42e0758b103b570","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:26:38.602285Z","signature_b64":"tXE/IcHIkDemhzf//75gIUUf+/cOtqspAYomd+0wFs/4zqj68qME44pKyXpIGYXIt0RgEkWyWzPETlAbLIwjBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0a7ed66f117a60f177a725ffcf693bb677b77301534a1885e42e0758b103b570","last_reissued_at":"2026-07-05T05:26:38.601801Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:26:38.601801Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2202.06539","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":"2026-07-05T05:26:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"IwiGnEIyAxf2t9j0HcvavqgSPcO5utBO/554yqNyQF8A+VD1NRGYB1WmOCjvRV9WSYqOfch26P6VZdY+olnLAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T21:14:26.182067Z"},"content_sha256":"9ac8ee095e40fcae6c9804a62a9a7c80537c65dab4f78772dd5f7275de06d89c","schema_version":"1.0","event_id":"sha256:9ac8ee095e40fcae6c9804a62a9a7c80537c65dab4f78772dd5f7275de06d89c"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2022:BJ7NM3YRPJQPC55HEX7462J3WZ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Deduplicating Training Data Mitigates Privacy Risks in Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.CR","authors_text":"Colin Raffel, Eric Wallace, Nikhil Kandpal","submitted_at":"2022-02-14T08:20:15Z","abstract_excerpt":"Past work has shown that large language models are susceptible to privacy attacks, where adversaries generate sequences from a trained model and detect which sequences are memorized from the training set. In this work, we show that the success of these attacks is largely due to duplication in commonly used web-scraped training sets. We first show that the rate at which language models regenerate training sequences is superlinearly related to a sequence's count in the training set. For instance, a sequence that is present 10 times in the training data is on average generated ~1000 times more of"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2202.06539","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/2202.06539/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":"2026-07-05T05:26:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xVHSGU/4JtYjwHRu0SCQ21N8l9JCpXQ1tMjTDEbLGTDrt4uAhimRiC0ejo4sb6DE79y8pffxYKcCMQ6SzkJwDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T21:14:26.182772Z"},"content_sha256":"1dab9acef4c0d755bd1ab3c94b3d09bc22531cc8a2265a82b2e439d26ab76fd1","schema_version":"1.0","event_id":"sha256:1dab9acef4c0d755bd1ab3c94b3d09bc22531cc8a2265a82b2e439d26ab76fd1"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/BJ7NM3YRPJQPC55HEX7462J3WZ/bundle.json","state_url":"https://pith.science/pith/BJ7NM3YRPJQPC55HEX7462J3WZ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/BJ7NM3YRPJQPC55HEX7462J3WZ/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":"2026-08-18T21:14:26Z","links":{"resolver":"https://pith.science/pith/BJ7NM3YRPJQPC55HEX7462J3WZ","bundle":"https://pith.science/pith/BJ7NM3YRPJQPC55HEX7462J3WZ/bundle.json","state":"https://pith.science/pith/BJ7NM3YRPJQPC55HEX7462J3WZ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/BJ7NM3YRPJQPC55HEX7462J3WZ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2022:BJ7NM3YRPJQPC55HEX7462J3WZ","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":"191f795825bd78e9a5fc70a749428eed511293417d27416a73301b531822d273","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2022-02-14T08:20:15Z","title_canon_sha256":"4e6d30a501403d539b164a84db0b93856b64a233ab359a29d18dce646972e21e"},"schema_version":"1.0","source":{"id":"2202.06539","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2202.06539","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"arxiv_version","alias_value":"2202.06539v3","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2202.06539","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"pith_short_12","alias_value":"BJ7NM3YRPJQP","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"pith_short_16","alias_value":"BJ7NM3YRPJQPC55H","created_at":"2026-07-05T05:26:38Z"},{"alias_kind":"pith_short_8","alias_value":"BJ7NM3YR","created_at":"2026-07-05T05:26:38Z"}],"graph_snapshots":[{"event_id":"sha256:1dab9acef4c0d755bd1ab3c94b3d09bc22531cc8a2265a82b2e439d26ab76fd1","target":"graph","created_at":"2026-07-05T05:26:38Z","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/2202.06539/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Past work has shown that large language models are susceptible to privacy attacks, where adversaries generate sequences from a trained model and detect which sequences are memorized from the training set. In this work, we show that the success of these attacks is largely due to duplication in commonly used web-scraped training sets. We first show that the rate at which language models regenerate training sequences is superlinearly related to a sequence's count in the training set. For instance, a sequence that is present 10 times in the training data is on average generated ~1000 times more of","authors_text":"Colin Raffel, Eric Wallace, Nikhil Kandpal","cross_cats":["cs.CL","cs.LG"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2022-02-14T08:20:15Z","title":"Deduplicating Training Data Mitigates Privacy Risks in Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2202.06539","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":"sha256:9ac8ee095e40fcae6c9804a62a9a7c80537c65dab4f78772dd5f7275de06d89c","target":"record","created_at":"2026-07-05T05:26:38Z","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":"191f795825bd78e9a5fc70a749428eed511293417d27416a73301b531822d273","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2022-02-14T08:20:15Z","title_canon_sha256":"4e6d30a501403d539b164a84db0b93856b64a233ab359a29d18dce646972e21e"},"schema_version":"1.0","source":{"id":"2202.06539","kind":"arxiv","version":3}},"canonical_sha256":"0a7ed66f117a60f177a725ffcf693bb677b77301534a1885e42e0758b103b570","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"0a7ed66f117a60f177a725ffcf693bb677b77301534a1885e42e0758b103b570","first_computed_at":"2026-07-05T05:26:38.601801Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T05:26:38.601801Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"tXE/IcHIkDemhzf//75gIUUf+/cOtqspAYomd+0wFs/4zqj68qME44pKyXpIGYXIt0RgEkWyWzPETlAbLIwjBw==","signature_status":"signed_v1","signed_at":"2026-07-05T05:26:38.602285Z","signed_message":"canonical_sha256_bytes"},"source_id":"2202.06539","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:9ac8ee095e40fcae6c9804a62a9a7c80537c65dab4f78772dd5f7275de06d89c","sha256:1dab9acef4c0d755bd1ab3c94b3d09bc22531cc8a2265a82b2e439d26ab76fd1"],"state_sha256":"45013bcd72318b8b2351f2cb394547552150bf6b6f939b5e65780de46ba2ea4c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"4LmP7EDfMyO1FIocpEQqjq4IWVGol2/mqvh1r6kj6DYIdXR6E6yS2VPFDn1tSOq74WjSXvlsXZDKTPlcHaeKDA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-18T21:14:26.189791Z","bundle_sha256":"224fcd43b2fb01539eff3ca7a42eb07bb438b8825de1e1788749bdf48a48ec68"}}