{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:URLCL3QBLADA52KRQ6MESMTKSS","short_pith_number":"pith:URLCL3QB","schema_version":"1.0","canonical_sha256":"a45625ee0158060ee951879849326a94a5d6e42e97d2ce81ab772c0eeea16911","source":{"kind":"arxiv","id":"2505.14633","version":1},"attestation_state":"computed","paper":{"title":"Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.HC","cs.LG"],"primary_cat":"cs.CL","authors_text":"Evan Hubinger, Kyle Fish, Sharan Maiya, Sydney Levine, Yejin Choi, Yu Ying Chiu, Zhilin Wang","submitted_at":"2025-05-20T17:24:09Z","abstract_excerpt":"Detecting AI risks becomes more challenging as stronger models emerge and find novel methods such as Alignment Faking to circumvent these detection attempts. Inspired by how risky behaviors in humans (i.e., illegal activities that may hurt others) are sometimes guided by strongly-held values, we believe that identifying values within AI models can be an early warning system for AI's risky behaviors. We create LitmusValues, an evaluation pipeline to reveal AI models' priorities on a range of AI value classes. Then, we collect AIRiskDilemmas, a diverse collection of dilemmas that pit values agai"},"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":"2505.14633","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-20T17:24:09Z","cross_cats_sorted":["cs.AI","cs.CY","cs.HC","cs.LG"],"title_canon_sha256":"d98a4f5d4ae78afb845da0cc799615168eba54630c134e113df5374e2a3372cf","abstract_canon_sha256":"8db253985a8c5ce438b958f6377c3cc3f0e62aa6880ce045f051cb4cffaba474"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:16.737559Z","signature_b64":"A5cQ8jtnk5j/4dkBsucwwNHo6KWnpcBhHRo+onCx7BQmedC5KGPdnOSrRwZtuMfDBln7jvWwHbuacwbmyIFRCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a45625ee0158060ee951879849326a94a5d6e42e97d2ce81ab772c0eeea16911","last_reissued_at":"2026-07-05T11:06:16.737035Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:16.737035Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Will AI Tell Lies to Save Sick Children? Litmus-Testing AI Values Prioritization with AIRiskDilemmas","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.HC","cs.LG"],"primary_cat":"cs.CL","authors_text":"Evan Hubinger, Kyle Fish, Sharan Maiya, Sydney Levine, Yejin Choi, Yu Ying Chiu, Zhilin Wang","submitted_at":"2025-05-20T17:24:09Z","abstract_excerpt":"Detecting AI risks becomes more challenging as stronger models emerge and find novel methods such as Alignment Faking to circumvent these detection attempts. Inspired by how risky behaviors in humans (i.e., illegal activities that may hurt others) are sometimes guided by strongly-held values, we believe that identifying values within AI models can be an early warning system for AI's risky behaviors. We create LitmusValues, an evaluation pipeline to reveal AI models' priorities on a range of AI value classes. Then, we collect AIRiskDilemmas, a diverse collection of dilemmas that pit values agai"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.14633","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/2505.14633/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":"2505.14633","created_at":"2026-07-05T11:06:16.737096+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.14633v1","created_at":"2026-07-05T11:06:16.737096+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.14633","created_at":"2026-07-05T11:06:16.737096+00:00"},{"alias_kind":"pith_short_12","alias_value":"URLCL3QBLADA","created_at":"2026-07-05T11:06:16.737096+00:00"},{"alias_kind":"pith_short_16","alias_value":"URLCL3QBLADA52KR","created_at":"2026-07-05T11:06:16.737096+00:00"},{"alias_kind":"pith_short_8","alias_value":"URLCL3QB","created_at":"2026-07-05T11:06:16.737096+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.09893","citing_title":"Pseudo-Deliberation in Language Models: When Reasoning Fails to Align Values and Actions","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21095","citing_title":"Backchaining Loss of Control Mitigations from Mission-Specific Benchmarks in National Security","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13339","citing_title":"Probing Persona-Dependent Preferences in Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09893","citing_title":"Pseudo-Deliberation in Language Models: When Reasoning Fails to Align Values and Actions","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21871","citing_title":"Machine Behavior in Relational Moral Dilemmas: Moral Rightness, Predicted Human Behavior, and Model Decisions","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS","json":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS.json","graph_json":"https://pith.science/api/pith-number/URLCL3QBLADA52KRQ6MESMTKSS/graph.json","events_json":"https://pith.science/api/pith-number/URLCL3QBLADA52KRQ6MESMTKSS/events.json","paper":"https://pith.science/paper/URLCL3QB"},"agent_actions":{"view_html":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS","download_json":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS.json","view_paper":"https://pith.science/paper/URLCL3QB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.14633&json=true","fetch_graph":"https://pith.science/api/pith-number/URLCL3QBLADA52KRQ6MESMTKSS/graph.json","fetch_events":"https://pith.science/api/pith-number/URLCL3QBLADA52KRQ6MESMTKSS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS/action/storage_attestation","attest_author":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS/action/author_attestation","sign_citation":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS/action/citation_signature","submit_replication":"https://pith.science/pith/URLCL3QBLADA52KRQ6MESMTKSS/action/replication_record"}},"created_at":"2026-07-05T11:06:16.737096+00:00","updated_at":"2026-07-05T11:06:16.737096+00:00"}