{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:AQLXV5MUNMZC3YH7GYOEBZJKCU","short_pith_number":"pith:AQLXV5MU","schema_version":"1.0","canonical_sha256":"04177af5946b322de0ff361c40e52a150888646833f38437552810a64d057c0d","source":{"kind":"arxiv","id":"2402.10877","version":7},"attestation_state":"computed","paper":{"title":"Robust agents learn causal world models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Jonathan Richens, Tom Everitt","submitted_at":"2024-02-16T18:29:19Z","abstract_excerpt":"It has long been hypothesised that causal reasoning plays a fundamental role in robust and general intelligence. However, it is not known if agents must learn causal models in order to generalise to new domains, or if other inductive biases are sufficient. We answer this question, showing that any agent capable of satisfying a regret bound under a large set of distributional shifts must have learned an approximate causal model of the data generating process, which converges to the true causal model for optimal agents. We discuss the implications of this result for several research areas includ"},"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":"2402.10877","kind":"arxiv","version":7},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-02-16T18:29:19Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"ae0a5229f5032381e88facacb88c6cc8f3ee0b2077a3e8486a2657db3dd90086","abstract_canon_sha256":"e159be954ae63b1b864190d4a33a2e6f001de39315d5030fd0add7909e0c25f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:45:50.059014Z","signature_b64":"YFrpFZ2IahnRB1bsLGFJZ4qbH3bxsD6p9T7JkifPIqxOCfC3PsiaKBNQJ8VQJmQxZMKI6fJXSCP+5bL2gRkwAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"04177af5946b322de0ff361c40e52a150888646833f38437552810a64d057c0d","last_reissued_at":"2026-07-05T08:45:50.058604Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:45:50.058604Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Robust agents learn causal world models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Jonathan Richens, Tom Everitt","submitted_at":"2024-02-16T18:29:19Z","abstract_excerpt":"It has long been hypothesised that causal reasoning plays a fundamental role in robust and general intelligence. However, it is not known if agents must learn causal models in order to generalise to new domains, or if other inductive biases are sufficient. We answer this question, showing that any agent capable of satisfying a regret bound under a large set of distributional shifts must have learned an approximate causal model of the data generating process, which converges to the true causal model for optimal agents. We discuss the implications of this result for several research areas includ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.10877","kind":"arxiv","version":7},"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/2402.10877/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":"2402.10877","created_at":"2026-07-05T08:45:50.058658+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.10877v7","created_at":"2026-07-05T08:45:50.058658+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.10877","created_at":"2026-07-05T08:45:50.058658+00:00"},{"alias_kind":"pith_short_12","alias_value":"AQLXV5MUNMZC","created_at":"2026-07-05T08:45:50.058658+00:00"},{"alias_kind":"pith_short_16","alias_value":"AQLXV5MUNMZC3YH7","created_at":"2026-07-05T08:45:50.058658+00:00"},{"alias_kind":"pith_short_8","alias_value":"AQLXV5MU","created_at":"2026-07-05T08:45:50.058658+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.29799","citing_title":"The CRISTAL Method: Neurosymbolic analysis from AI-synthesized world models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27580","citing_title":"You Are in Control of Your State: Why Human Outcomes Are Controllable Through Causal State Intervention","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02681","citing_title":"The Design and Composition of Structural Causal Decision Processes","ref_index":44,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU","json":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU.json","graph_json":"https://pith.science/api/pith-number/AQLXV5MUNMZC3YH7GYOEBZJKCU/graph.json","events_json":"https://pith.science/api/pith-number/AQLXV5MUNMZC3YH7GYOEBZJKCU/events.json","paper":"https://pith.science/paper/AQLXV5MU"},"agent_actions":{"view_html":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU","download_json":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU.json","view_paper":"https://pith.science/paper/AQLXV5MU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.10877&json=true","fetch_graph":"https://pith.science/api/pith-number/AQLXV5MUNMZC3YH7GYOEBZJKCU/graph.json","fetch_events":"https://pith.science/api/pith-number/AQLXV5MUNMZC3YH7GYOEBZJKCU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU/action/storage_attestation","attest_author":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU/action/author_attestation","sign_citation":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU/action/citation_signature","submit_replication":"https://pith.science/pith/AQLXV5MUNMZC3YH7GYOEBZJKCU/action/replication_record"}},"created_at":"2026-07-05T08:45:50.058658+00:00","updated_at":"2026-07-05T08:45:50.058658+00:00"}