{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WBAKENOTLDYXEBCSLEPZBVYLY2","short_pith_number":"pith:WBAKENOT","schema_version":"1.0","canonical_sha256":"b040a235d358f1720452591f90d70bc68ec1c167e7aa2cb98f9cf6258641b6f9","source":{"kind":"arxiv","id":"2404.00486","version":1},"attestation_state":"computed","paper":{"title":"Dialectical Alignment: Resolving the Tension of 3H and Security Threats of LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Di Wang, Han Jiang, Jiayuan Su, Keyuan Cheng, Lijie Hu, Mengdi Li, Muhammad Asif Ali, Shu Yang","submitted_at":"2024-03-30T22:41:05Z","abstract_excerpt":"With the rise of large language models (LLMs), ensuring they embody the principles of being helpful, honest, and harmless (3H), known as Human Alignment, becomes crucial. While existing alignment methods like RLHF, DPO, etc., effectively fine-tune LLMs to match preferences in the preference dataset, they often lead LLMs to highly receptive human input and external evidence, even when this information is poisoned. This leads to a tendency for LLMs to be Adaptive Chameleons when external evidence conflicts with their parametric memory. This exacerbates the risk of LLM being attacked by external "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2404.00486","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-03-30T22:41:05Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"aa4308400601c13a5cdf97eb668221d3998737ac84e185a9f3a333af97d6bb0d","abstract_canon_sha256":"b64ef4cdf2d40f0edf44373fb2fe2b36b2270288f0b320da99ec11dc6621dbf7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:02:34.064022Z","signature_b64":"zOh2zUnElwFxXt9ztxNZ6/tfDdrYwIJUI1yr2rgsZAI1B755P3lkf8+Ux+bPV1Cd7lkpzEavLo7u7icm5qAqBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b040a235d358f1720452591f90d70bc68ec1c167e7aa2cb98f9cf6258641b6f9","last_reissued_at":"2026-07-05T08:02:34.062666Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:02:34.062666Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dialectical Alignment: Resolving the Tension of 3H and Security Threats of LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Di Wang, Han Jiang, Jiayuan Su, Keyuan Cheng, Lijie Hu, Mengdi Li, Muhammad Asif Ali, Shu Yang","submitted_at":"2024-03-30T22:41:05Z","abstract_excerpt":"With the rise of large language models (LLMs), ensuring they embody the principles of being helpful, honest, and harmless (3H), known as Human Alignment, becomes crucial. While existing alignment methods like RLHF, DPO, etc., effectively fine-tune LLMs to match preferences in the preference dataset, they often lead LLMs to highly receptive human input and external evidence, even when this information is poisoned. This leads to a tendency for LLMs to be Adaptive Chameleons when external evidence conflicts with their parametric memory. This exacerbates the risk of LLM being attacked by external "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.00486","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/2404.00486/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2404.00486","created_at":"2026-07-05T08:02:34.062729+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.00486v1","created_at":"2026-07-05T08:02:34.062729+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.00486","created_at":"2026-07-05T08:02:34.062729+00:00"},{"alias_kind":"pith_short_12","alias_value":"WBAKENOTLDYX","created_at":"2026-07-05T08:02:34.062729+00:00"},{"alias_kind":"pith_short_16","alias_value":"WBAKENOTLDYXEBCS","created_at":"2026-07-05T08:02:34.062729+00:00"},{"alias_kind":"pith_short_8","alias_value":"WBAKENOT","created_at":"2026-07-05T08:02:34.062729+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2","json":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2.json","graph_json":"https://pith.science/api/pith-number/WBAKENOTLDYXEBCSLEPZBVYLY2/graph.json","events_json":"https://pith.science/api/pith-number/WBAKENOTLDYXEBCSLEPZBVYLY2/events.json","paper":"https://pith.science/paper/WBAKENOT"},"agent_actions":{"view_html":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2","download_json":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2.json","view_paper":"https://pith.science/paper/WBAKENOT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.00486&json=true","fetch_graph":"https://pith.science/api/pith-number/WBAKENOTLDYXEBCSLEPZBVYLY2/graph.json","fetch_events":"https://pith.science/api/pith-number/WBAKENOTLDYXEBCSLEPZBVYLY2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2/action/storage_attestation","attest_author":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2/action/author_attestation","sign_citation":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2/action/citation_signature","submit_replication":"https://pith.science/pith/WBAKENOTLDYXEBCSLEPZBVYLY2/action/replication_record"}},"created_at":"2026-07-05T08:02:34.062729+00:00","updated_at":"2026-07-05T08:02:34.062729+00:00"}