{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:SPVIOYDQDU6RF33WFH2BXJGQ5L","short_pith_number":"pith:SPVIOYDQ","schema_version":"1.0","canonical_sha256":"93ea8760701d3d12ef7629f41ba4d0ead9a61bb58afbd91a706d116f41f646e3","source":{"kind":"arxiv","id":"2605.24229","version":1},"attestation_state":"computed","paper":{"title":"How Well Do Models Follow Their Constitutions?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Arya Jakkli, Neel Nanda, Senthooran Rajamanoharan","submitted_at":"2026-05-22T21:17:16Z","abstract_excerpt":"Frontier AI developers now train models against long written behavioral specifications, such as Anthropic's constitution (Anthropic, 2025a) and OpenAI's Model Spec (OpenAI, 2025a), integrated into post-training via methods like character training (Anthropic, 2024) and deliberative alignment (Guan et al., 2024). These documents serve a governance function, but it is unclear how well models actually follow them under adversarial, multi-turn pressure similar to what they would face in real-world deployment. We propose a multi-method audit pipeline that treats each lab's published specification as"},"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":"2605.24229","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-22T21:17:16Z","cross_cats_sorted":[],"title_canon_sha256":"83218a263c270e43982eecf91161a404750973d6a8a539b6d66ded5ddf128aea","abstract_canon_sha256":"a608e468b679e9344da0c195db8131091d02ba6226b1e0347e0a125336b9bfb7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:02:53.887847Z","signature_b64":"Q10Uha6i1604HsLaxRfBo2F5O9MYbAWRzz75lii7q8s5WOjN3FZcU2rJtKOhtXLTgfzjc7/+dC2twGgNGdVMBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"93ea8760701d3d12ef7629f41ba4d0ead9a61bb58afbd91a706d116f41f646e3","last_reissued_at":"2026-05-26T01:02:53.887180Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:02:53.887180Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Well Do Models Follow Their Constitutions?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Arya Jakkli, Neel Nanda, Senthooran Rajamanoharan","submitted_at":"2026-05-22T21:17:16Z","abstract_excerpt":"Frontier AI developers now train models against long written behavioral specifications, such as Anthropic's constitution (Anthropic, 2025a) and OpenAI's Model Spec (OpenAI, 2025a), integrated into post-training via methods like character training (Anthropic, 2024) and deliberative alignment (Guan et al., 2024). These documents serve a governance function, but it is unclear how well models actually follow them under adversarial, multi-turn pressure similar to what they would face in real-world deployment. We propose a multi-method audit pipeline that treats each lab's published specification as"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.24229","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/2605.24229/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":"2605.24229","created_at":"2026-05-26T01:02:53.887273+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.24229v1","created_at":"2026-05-26T01:02:53.887273+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.24229","created_at":"2026-05-26T01:02:53.887273+00:00"},{"alias_kind":"pith_short_12","alias_value":"SPVIOYDQDU6R","created_at":"2026-05-26T01:02:53.887273+00:00"},{"alias_kind":"pith_short_16","alias_value":"SPVIOYDQDU6RF33W","created_at":"2026-05-26T01:02:53.887273+00:00"},{"alias_kind":"pith_short_8","alias_value":"SPVIOYDQ","created_at":"2026-05-26T01:02:53.887273+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/SPVIOYDQDU6RF33WFH2BXJGQ5L","json":"https://pith.science/pith/SPVIOYDQDU6RF33WFH2BXJGQ5L.json","graph_json":"https://pith.science/api/pith-number/SPVIOYDQDU6RF33WFH2BXJGQ5L/graph.json","events_json":"https://pith.science/api/pith-number/SPVIOYDQDU6RF33WFH2BXJGQ5L/events.json","paper":"https://pith.science/paper/SPVIOYDQ"},"agent_actions":{"view_html":"https://pith.science/pith/SPVIOYDQDU6RF33WFH2BXJGQ5L","download_json":"https://pith.science/pith/SPVIOYDQDU6RF33WFH2BXJGQ5L.json","view_paper":"https://pith.science/paper/SPVIOYDQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.24229&json=true","fetch_graph":"https://pith.science/api/pith-number/SPVIOYDQDU6RF33WFH2BXJGQ5L/graph.json","fetch_events":"https://pith.science/api/pith-number/SPVIOYDQDU6RF33WFH2BXJGQ5L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SPVIOYDQDU6RF33WFH2BXJGQ5L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SPVIOYDQDU6RF33WFH2BXJGQ5L/action/storage_attestation","attest_author":"https://pith.science/pith/SPVIOYDQDU6RF33WFH2BXJGQ5L/action/author_attestation","sign_citation":"https://pith.science/pith/SPVIOYDQDU6RF33WFH2BXJGQ5L/action/citation_signature","submit_replication":"https://pith.science/pith/SPVIOYDQDU6RF33WFH2BXJGQ5L/action/replication_record"}},"created_at":"2026-05-26T01:02:53.887273+00:00","updated_at":"2026-05-26T01:02:53.887273+00:00"}