{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:G3FPC7243ZVK2AACD3I7VUMYAA","short_pith_number":"pith:G3FPC724","schema_version":"1.0","canonical_sha256":"36caf17f5cde6aad00021ed1fad1980009066f95d496783aa9944a36005e1367","source":{"kind":"arxiv","id":"2407.15286","version":3},"attestation_state":"computed","paper":{"title":"Intrinsic Self-correction for Enhanced Morality: An Analysis of Internal Mechanisms and the Superficial Hypothesis","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Guangliang Liu, Haitao Mao, Jiliang Tang, Kristen Marie Johnson","submitted_at":"2024-07-21T22:50:11Z","abstract_excerpt":"Large Language Models (LLMs) are capable of producing content that perpetuates stereotypes, discrimination, and toxicity. The recently proposed moral self-correction is a computationally efficient method for reducing harmful content in the responses of LLMs. However, the process of how injecting self-correction instructions can modify the behavior of LLMs remains under-explored. In this paper, we explore the effectiveness of moral self-correction by answering three research questions: (1) In what scenarios does moral self-correction work? (2) What are the internal mechanisms of LLMs, e.g., hid"},"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":"2407.15286","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-21T22:50:11Z","cross_cats_sorted":[],"title_canon_sha256":"c4c74a7bab17b29fcb8c3600147e34b7e570668d7fd1f0b8a6271823cf0be834","abstract_canon_sha256":"c27f1f6bcdadef19657c2e4e092faca536ce99691368ad4abaaab98a8f55ea81"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:17:14.703441Z","signature_b64":"EFq2QVDTeKhU5brrdao7/7VF4UAarghgZVHDWgkKY/rNJeHDctwGPtXVKxv3EKHSwD/iVLznx6MDhbk5Gw1wAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"36caf17f5cde6aad00021ed1fad1980009066f95d496783aa9944a36005e1367","last_reissued_at":"2026-07-05T09:17:14.703020Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:17:14.703020Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Intrinsic Self-correction for Enhanced Morality: An Analysis of Internal Mechanisms and the Superficial Hypothesis","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Guangliang Liu, Haitao Mao, Jiliang Tang, Kristen Marie Johnson","submitted_at":"2024-07-21T22:50:11Z","abstract_excerpt":"Large Language Models (LLMs) are capable of producing content that perpetuates stereotypes, discrimination, and toxicity. The recently proposed moral self-correction is a computationally efficient method for reducing harmful content in the responses of LLMs. However, the process of how injecting self-correction instructions can modify the behavior of LLMs remains under-explored. In this paper, we explore the effectiveness of moral self-correction by answering three research questions: (1) In what scenarios does moral self-correction work? (2) What are the internal mechanisms of LLMs, e.g., hid"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.15286","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/2407.15286/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":"2407.15286","created_at":"2026-07-05T09:17:14.703076+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.15286v3","created_at":"2026-07-05T09:17:14.703076+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.15286","created_at":"2026-07-05T09:17:14.703076+00:00"},{"alias_kind":"pith_short_12","alias_value":"G3FPC7243ZVK","created_at":"2026-07-05T09:17:14.703076+00:00"},{"alias_kind":"pith_short_16","alias_value":"G3FPC7243ZVK2AAC","created_at":"2026-07-05T09:17:14.703076+00:00"},{"alias_kind":"pith_short_8","alias_value":"G3FPC724","created_at":"2026-07-05T09:17:14.703076+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.05282","citing_title":"Not All Errors Are Created Equal: ASCoT Addresses Late-Stage Fragility in Efficient LLM Reasoning","ref_index":32,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA","json":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA.json","graph_json":"https://pith.science/api/pith-number/G3FPC7243ZVK2AACD3I7VUMYAA/graph.json","events_json":"https://pith.science/api/pith-number/G3FPC7243ZVK2AACD3I7VUMYAA/events.json","paper":"https://pith.science/paper/G3FPC724"},"agent_actions":{"view_html":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA","download_json":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA.json","view_paper":"https://pith.science/paper/G3FPC724","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.15286&json=true","fetch_graph":"https://pith.science/api/pith-number/G3FPC7243ZVK2AACD3I7VUMYAA/graph.json","fetch_events":"https://pith.science/api/pith-number/G3FPC7243ZVK2AACD3I7VUMYAA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA/action/storage_attestation","attest_author":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA/action/author_attestation","sign_citation":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA/action/citation_signature","submit_replication":"https://pith.science/pith/G3FPC7243ZVK2AACD3I7VUMYAA/action/replication_record"}},"created_at":"2026-07-05T09:17:14.703076+00:00","updated_at":"2026-07-05T09:17:14.703076+00:00"}