{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:F3B3KGLJNHD7VAXVZBMTULONZY","short_pith_number":"pith:F3B3KGLJ","schema_version":"1.0","canonical_sha256":"2ec3b5196969c7fa82f5c8593a2dcdce2f994f11ec8ba226ab9f89602abfcd13","source":{"kind":"arxiv","id":"2503.17365","version":2},"attestation_state":"computed","paper":{"title":"How Effective Is Constitutional AI in Small LLMs? A Study on DeepSeek-R1 and Its Peers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY"],"primary_cat":"cs.LG","authors_text":"Antonio-Gabriel Chac\\'on Menke (Shibaura Institute of Technology, Kempten University of Applied Sciences), Phan Xuan Tan (Shibaura Institute of Technology)","submitted_at":"2025-02-01T06:58:27Z","abstract_excerpt":"Recent incidents highlight safety risks in Large Language Models (LLMs), motivating research into alignment methods like Constitutional AI (CAI). This paper explores CAI's self-critique mechanism on small, uncensored 7-9B parameter models: DeepSeek-R1-8B, Gemma-2-9B, Llama 3.1-8B, and Qwen2.5-7B. We show that while Llama-based models exhibited significant harm reduction through self-critique, other architectures demonstrated less improvement in harm detection after abliteration. These results suggest CAI's effectiveness may vary depending on model architecture and reasoning capabilities."},"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":"2503.17365","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-01T06:58:27Z","cross_cats_sorted":["cs.AI","cs.CY"],"title_canon_sha256":"a77de42ad99865405473f147e356cd53ea0975f9a8d9c94b04bab99e6b16d3c3","abstract_canon_sha256":"5153960960df084090a6a0456a9608c65b18f89d6e021a7cd1db348c475e1c28"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:47:37.033708Z","signature_b64":"MiXUDZbD7+9VjA2N1i1/LKKH3FMeCXfjyhUX8vi91Sp42ISoMaQjvFvejW2LRX7a0GBr7b2ivGT2heF7hL8NAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2ec3b5196969c7fa82f5c8593a2dcdce2f994f11ec8ba226ab9f89602abfcd13","last_reissued_at":"2026-07-05T10:47:37.033173Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:47:37.033173Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Effective Is Constitutional AI in Small LLMs? A Study on DeepSeek-R1 and Its Peers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY"],"primary_cat":"cs.LG","authors_text":"Antonio-Gabriel Chac\\'on Menke (Shibaura Institute of Technology, Kempten University of Applied Sciences), Phan Xuan Tan (Shibaura Institute of Technology)","submitted_at":"2025-02-01T06:58:27Z","abstract_excerpt":"Recent incidents highlight safety risks in Large Language Models (LLMs), motivating research into alignment methods like Constitutional AI (CAI). This paper explores CAI's self-critique mechanism on small, uncensored 7-9B parameter models: DeepSeek-R1-8B, Gemma-2-9B, Llama 3.1-8B, and Qwen2.5-7B. We show that while Llama-based models exhibited significant harm reduction through self-critique, other architectures demonstrated less improvement in harm detection after abliteration. These results suggest CAI's effectiveness may vary depending on model architecture and reasoning capabilities."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.17365","kind":"arxiv","version":2},"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/2503.17365/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":"2503.17365","created_at":"2026-07-05T10:47:37.033229+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.17365v2","created_at":"2026-07-05T10:47:37.033229+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.17365","created_at":"2026-07-05T10:47:37.033229+00:00"},{"alias_kind":"pith_short_12","alias_value":"F3B3KGLJNHD7","created_at":"2026-07-05T10:47:37.033229+00:00"},{"alias_kind":"pith_short_16","alias_value":"F3B3KGLJNHD7VAXV","created_at":"2026-07-05T10:47:37.033229+00:00"},{"alias_kind":"pith_short_8","alias_value":"F3B3KGLJ","created_at":"2026-07-05T10:47:37.033229+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/F3B3KGLJNHD7VAXVZBMTULONZY","json":"https://pith.science/pith/F3B3KGLJNHD7VAXVZBMTULONZY.json","graph_json":"https://pith.science/api/pith-number/F3B3KGLJNHD7VAXVZBMTULONZY/graph.json","events_json":"https://pith.science/api/pith-number/F3B3KGLJNHD7VAXVZBMTULONZY/events.json","paper":"https://pith.science/paper/F3B3KGLJ"},"agent_actions":{"view_html":"https://pith.science/pith/F3B3KGLJNHD7VAXVZBMTULONZY","download_json":"https://pith.science/pith/F3B3KGLJNHD7VAXVZBMTULONZY.json","view_paper":"https://pith.science/paper/F3B3KGLJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.17365&json=true","fetch_graph":"https://pith.science/api/pith-number/F3B3KGLJNHD7VAXVZBMTULONZY/graph.json","fetch_events":"https://pith.science/api/pith-number/F3B3KGLJNHD7VAXVZBMTULONZY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F3B3KGLJNHD7VAXVZBMTULONZY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F3B3KGLJNHD7VAXVZBMTULONZY/action/storage_attestation","attest_author":"https://pith.science/pith/F3B3KGLJNHD7VAXVZBMTULONZY/action/author_attestation","sign_citation":"https://pith.science/pith/F3B3KGLJNHD7VAXVZBMTULONZY/action/citation_signature","submit_replication":"https://pith.science/pith/F3B3KGLJNHD7VAXVZBMTULONZY/action/replication_record"}},"created_at":"2026-07-05T10:47:37.033229+00:00","updated_at":"2026-07-05T10:47:37.033229+00:00"}