{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:PLHULOOTPKS7MU37QWJXIS22G7","short_pith_number":"pith:PLHULOOT","canonical_record":{"source":{"id":"2504.05050","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-07T13:20:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8050f057433029e626ca76438c59e1febbb2b153f34ecf5af7ff7a5141e2ef41","abstract_canon_sha256":"268b4eb1790c5f9191c24f30008dd3245103fbd74530576d7fc68c3bdc224bda"},"schema_version":"1.0"},"canonical_sha256":"7acf45b9d37aa5f6537f8593744b5a37fa174e38a697de94cbd56587cfc71d33","source":{"kind":"arxiv","id":"2504.05050","version":4},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.05050","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"arxiv_version","alias_value":"2504.05050v4","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.05050","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"pith_short_12","alias_value":"PLHULOOTPKS7","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"pith_short_16","alias_value":"PLHULOOTPKS7MU37","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"pith_short_8","alias_value":"PLHULOOT","created_at":"2026-07-05T11:14:28Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:PLHULOOTPKS7MU37QWJXIS22G7","target":"record","payload":{"canonical_record":{"source":{"id":"2504.05050","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-07T13:20:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8050f057433029e626ca76438c59e1febbb2b153f34ecf5af7ff7a5141e2ef41","abstract_canon_sha256":"268b4eb1790c5f9191c24f30008dd3245103fbd74530576d7fc68c3bdc224bda"},"schema_version":"1.0"},"canonical_sha256":"7acf45b9d37aa5f6537f8593744b5a37fa174e38a697de94cbd56587cfc71d33","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:14:28.543985Z","signature_b64":"V2U2+JHk5GwQnt2NA1nUYWQXsFxcZFwuvJrHjUvLuGhFgaMpy6Agtj1d5FndBorXcgsYwRP8Z5jwB+2GbFl9Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7acf45b9d37aa5f6537f8593744b5a37fa174e38a697de94cbd56587cfc71d33","last_reissued_at":"2026-07-05T11:14:28.543473Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:14:28.543473Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2504.05050","source_version":4,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:14:28Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AobzQEsqbn5Kp3pgHl2kXx4JhO/Xfg7Rdf/L+ckyZqWf+DBxCRbssGQr5Jig9nn2iFYPKDmFtCPwuvpZcxtLAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T09:09:09.576991Z"},"content_sha256":"8f4ee499ccbbca28f98bedd02cb184810a4bf25966f85bf776a33de3e2b5f588","schema_version":"1.0","event_id":"sha256:8f4ee499ccbbca28f98bedd02cb184810a4bf25966f85bf776a33de3e2b5f588"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:PLHULOOTPKS7MU37QWJXIS22G7","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Revealing the Intrinsic Ethical Vulnerability of Aligned Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jianhong Pan, Jiawei Lian, Lap-Pui Chau, Lefan Wang, Shaohui Mei, Yi Wang","submitted_at":"2025-04-07T13:20:17Z","abstract_excerpt":"Large language models (LLMs) are foundational explorations to artificial general intelligence, yet their alignment with human values via instruction tuning and preference learning achieves only superficial compliance. Here, we demonstrate that harmful knowledge embedded during pretraining persists as indelible \"dark patterns\" in LLMs' parametric memory, evading alignment safeguards and resurfacing under adversarial inducement at distributional shifts. In this study, we first theoretically analyze the intrinsic ethical vulnerability of aligned LLMs by proving that current alignment methods yiel"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.05050","kind":"arxiv","version":4},"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/2504.05050/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":"2026-07-05T11:14:28Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qiYXAgShqE62nGv6gAE2c9uIhEahm+JxbjvJW6eBlMShEIQfQUjHTTk0IbIVdNGO++SOaj7SbPqlwXlZVuPWBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T09:09:09.577924Z"},"content_sha256":"bd149c1d1f57190bb6485e36b718e05b688c51a14936d85466a53e77c0e55ca4","schema_version":"1.0","event_id":"sha256:bd149c1d1f57190bb6485e36b718e05b688c51a14936d85466a53e77c0e55ca4"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/PLHULOOTPKS7MU37QWJXIS22G7/bundle.json","state_url":"https://pith.science/pith/PLHULOOTPKS7MU37QWJXIS22G7/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/PLHULOOTPKS7MU37QWJXIS22G7/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":"2026-08-10T09:09:09Z","links":{"resolver":"https://pith.science/pith/PLHULOOTPKS7MU37QWJXIS22G7","bundle":"https://pith.science/pith/PLHULOOTPKS7MU37QWJXIS22G7/bundle.json","state":"https://pith.science/pith/PLHULOOTPKS7MU37QWJXIS22G7/state.json","well_known_bundle":"https://pith.science/.well-known/pith/PLHULOOTPKS7MU37QWJXIS22G7/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:PLHULOOTPKS7MU37QWJXIS22G7","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":"268b4eb1790c5f9191c24f30008dd3245103fbd74530576d7fc68c3bdc224bda","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-07T13:20:17Z","title_canon_sha256":"8050f057433029e626ca76438c59e1febbb2b153f34ecf5af7ff7a5141e2ef41"},"schema_version":"1.0","source":{"id":"2504.05050","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.05050","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"arxiv_version","alias_value":"2504.05050v4","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.05050","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"pith_short_12","alias_value":"PLHULOOTPKS7","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"pith_short_16","alias_value":"PLHULOOTPKS7MU37","created_at":"2026-07-05T11:14:28Z"},{"alias_kind":"pith_short_8","alias_value":"PLHULOOT","created_at":"2026-07-05T11:14:28Z"}],"graph_snapshots":[{"event_id":"sha256:bd149c1d1f57190bb6485e36b718e05b688c51a14936d85466a53e77c0e55ca4","target":"graph","created_at":"2026-07-05T11:14:28Z","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/2504.05050/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models (LLMs) are foundational explorations to artificial general intelligence, yet their alignment with human values via instruction tuning and preference learning achieves only superficial compliance. Here, we demonstrate that harmful knowledge embedded during pretraining persists as indelible \"dark patterns\" in LLMs' parametric memory, evading alignment safeguards and resurfacing under adversarial inducement at distributional shifts. In this study, we first theoretically analyze the intrinsic ethical vulnerability of aligned LLMs by proving that current alignment methods yiel","authors_text":"Jianhong Pan, Jiawei Lian, Lap-Pui Chau, Lefan Wang, Shaohui Mei, Yi Wang","cross_cats":["cs.AI"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-07T13:20:17Z","title":"Revealing the Intrinsic Ethical Vulnerability of Aligned Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.05050","kind":"arxiv","version":4},"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":"sha256:8f4ee499ccbbca28f98bedd02cb184810a4bf25966f85bf776a33de3e2b5f588","target":"record","created_at":"2026-07-05T11:14:28Z","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":"268b4eb1790c5f9191c24f30008dd3245103fbd74530576d7fc68c3bdc224bda","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-07T13:20:17Z","title_canon_sha256":"8050f057433029e626ca76438c59e1febbb2b153f34ecf5af7ff7a5141e2ef41"},"schema_version":"1.0","source":{"id":"2504.05050","kind":"arxiv","version":4}},"canonical_sha256":"7acf45b9d37aa5f6537f8593744b5a37fa174e38a697de94cbd56587cfc71d33","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"7acf45b9d37aa5f6537f8593744b5a37fa174e38a697de94cbd56587cfc71d33","first_computed_at":"2026-07-05T11:14:28.543473Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:14:28.543473Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"V2U2+JHk5GwQnt2NA1nUYWQXsFxcZFwuvJrHjUvLuGhFgaMpy6Agtj1d5FndBorXcgsYwRP8Z5jwB+2GbFl9Aw==","signature_status":"signed_v1","signed_at":"2026-07-05T11:14:28.543985Z","signed_message":"canonical_sha256_bytes"},"source_id":"2504.05050","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:8f4ee499ccbbca28f98bedd02cb184810a4bf25966f85bf776a33de3e2b5f588","sha256:bd149c1d1f57190bb6485e36b718e05b688c51a14936d85466a53e77c0e55ca4"],"state_sha256":"19e28087dab28a2d8001e263419648f20a8f2731eedce8c638a8e54cc4d25dde"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HOFSvjj2gHkqBCeyW7TcheRkZlmz+Pf7Hs7V5u3BE6zNuuLqOfYLeV4ZQ4WxTMS4T4jHykEADKMDUPjpbveqBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T09:09:09.583797Z","bundle_sha256":"a09b018ecc18684c26fd0af2bc11f36e7f12d806b9b1f508d3cb9e881f0aaa3a"}}