{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:4HMJG45RTUDCFKZ4OEYHLHSWBT","short_pith_number":"pith:4HMJG45R","schema_version":"1.0","canonical_sha256":"e1d89373b19d0622ab3c7130759e560cfff46454cd542217d8f6211750f6c2bf","source":{"kind":"arxiv","id":"2304.06528","version":1},"attestation_state":"computed","paper":{"title":"Power-seeking can be probable and predictive for trained agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Janos Kramar, Victoria Krakovna","submitted_at":"2023-04-13T13:29:01Z","abstract_excerpt":"Power-seeking behavior is a key source of risk from advanced AI, but our theoretical understanding of this phenomenon is relatively limited. Building on existing theoretical results demonstrating power-seeking incentives for most reward functions, we investigate how the training process affects power-seeking incentives and show that they are still likely to hold for trained agents under some simplifying assumptions. We formally define the training-compatible goal set (the set of goals consistent with the training rewards) and assume that the trained agent learns a goal from this set. In a sett"},"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":"2304.06528","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2023-04-13T13:29:01Z","cross_cats_sorted":[],"title_canon_sha256":"0ef07879454f6758976cb9d61e07af494b489fc93e0602c2f013465a8359502b","abstract_canon_sha256":"ba317ffffa9ff35e026952337c829558ff208f3e6193ab17eb481a6d690d41f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:00:44.896936Z","signature_b64":"Su7KnlRA9oYwNnRVJqlpGDLY7b27f1HALVQcrjs37no2HGr63bhJq4yTfk3zK4I2vDvZyPGL7KJaY7P5wiHjDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e1d89373b19d0622ab3c7130759e560cfff46454cd542217d8f6211750f6c2bf","last_reissued_at":"2026-07-05T06:00:44.896435Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:00:44.896435Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Power-seeking can be probable and predictive for trained agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Janos Kramar, Victoria Krakovna","submitted_at":"2023-04-13T13:29:01Z","abstract_excerpt":"Power-seeking behavior is a key source of risk from advanced AI, but our theoretical understanding of this phenomenon is relatively limited. Building on existing theoretical results demonstrating power-seeking incentives for most reward functions, we investigate how the training process affects power-seeking incentives and show that they are still likely to hold for trained agents under some simplifying assumptions. We formally define the training-compatible goal set (the set of goals consistent with the training rewards) and assume that the trained agent learns a goal from this set. In a sett"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2304.06528","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/2304.06528/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":"2304.06528","created_at":"2026-07-05T06:00:44.896480+00:00"},{"alias_kind":"arxiv_version","alias_value":"2304.06528v1","created_at":"2026-07-05T06:00:44.896480+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.06528","created_at":"2026-07-05T06:00:44.896480+00:00"},{"alias_kind":"pith_short_12","alias_value":"4HMJG45RTUDC","created_at":"2026-07-05T06:00:44.896480+00:00"},{"alias_kind":"pith_short_16","alias_value":"4HMJG45RTUDCFKZ4","created_at":"2026-07-05T06:00:44.896480+00:00"},{"alias_kind":"pith_short_8","alias_value":"4HMJG45R","created_at":"2026-07-05T06:00:44.896480+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.23094","citing_title":"Cognitive Digital Twins: Ethical Risks and Governance for AI Systems That Model the Mind","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30666","citing_title":"Reframing AGI Confrontation with Off Earth Autonomy","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT","json":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT.json","graph_json":"https://pith.science/api/pith-number/4HMJG45RTUDCFKZ4OEYHLHSWBT/graph.json","events_json":"https://pith.science/api/pith-number/4HMJG45RTUDCFKZ4OEYHLHSWBT/events.json","paper":"https://pith.science/paper/4HMJG45R"},"agent_actions":{"view_html":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT","download_json":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT.json","view_paper":"https://pith.science/paper/4HMJG45R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2304.06528&json=true","fetch_graph":"https://pith.science/api/pith-number/4HMJG45RTUDCFKZ4OEYHLHSWBT/graph.json","fetch_events":"https://pith.science/api/pith-number/4HMJG45RTUDCFKZ4OEYHLHSWBT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT/action/storage_attestation","attest_author":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT/action/author_attestation","sign_citation":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT/action/citation_signature","submit_replication":"https://pith.science/pith/4HMJG45RTUDCFKZ4OEYHLHSWBT/action/replication_record"}},"created_at":"2026-07-05T06:00:44.896480+00:00","updated_at":"2026-07-05T06:00:44.896480+00:00"}