{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:LDOSFEOGVFXXJI5HS57ZRTOYXC","short_pith_number":"pith:LDOSFEOG","schema_version":"1.0","canonical_sha256":"58dd2291c6a96f74a3a7977f98cdd8b8b6bbf3a5f7a0b8a815d6668fca83005c","source":{"kind":"arxiv","id":"1907.01475","version":1},"attestation_state":"computed","paper":{"title":"Generalizing from a few environments in safety-critical reinforcement learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Angelos Filos, Owain Evans, Yarin Gal, Zachary Kenton","submitted_at":"2019-07-02T16:12:34Z","abstract_excerpt":"Before deploying autonomous agents in the real world, we need to be confident they will perform safely in novel situations. Ideally, we would expose agents to a very wide range of situations during training, allowing them to learn about every possible danger, but this is often impractical. This paper investigates safety and generalization from a limited number of training environments in deep reinforcement learning (RL). We find RL algorithms can fail dangerously on unseen test environments even when performing perfectly on training environments. Firstly, in a gridworld setting, we show that c"},"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":"1907.01475","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-07-02T16:12:34Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"59349a745f4160954b0c1e9573b65e6fdf684c109574c151d11ef935425d1576","abstract_canon_sha256":"44bdf09742f1cd8ca8cc89e0ca8b63007461d6eaa4a6994bcc45e3e50b66180f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:41:40.297808Z","signature_b64":"hsqAIPCfXPD59tb4ARCQf80r2TyAQ/Z28JBnOpIJhNwEG30MSQzNqSlg3G4nHpOFPBJCQH++6oVocJQ/jQA+BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"58dd2291c6a96f74a3a7977f98cdd8b8b6bbf3a5f7a0b8a815d6668fca83005c","last_reissued_at":"2026-05-17T23:41:40.297328Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:41:40.297328Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Generalizing from a few environments in safety-critical reinforcement learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Angelos Filos, Owain Evans, Yarin Gal, Zachary Kenton","submitted_at":"2019-07-02T16:12:34Z","abstract_excerpt":"Before deploying autonomous agents in the real world, we need to be confident they will perform safely in novel situations. Ideally, we would expose agents to a very wide range of situations during training, allowing them to learn about every possible danger, but this is often impractical. This paper investigates safety and generalization from a limited number of training environments in deep reinforcement learning (RL). We find RL algorithms can fail dangerously on unseen test environments even when performing perfectly on training environments. Firstly, in a gridworld setting, we show that c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1907.01475","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":""},"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":"1907.01475","created_at":"2026-05-17T23:41:40.297395+00:00"},{"alias_kind":"arxiv_version","alias_value":"1907.01475v1","created_at":"2026-05-17T23:41:40.297395+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1907.01475","created_at":"2026-05-17T23:41:40.297395+00:00"},{"alias_kind":"pith_short_12","alias_value":"LDOSFEOGVFXX","created_at":"2026-05-18T12:33:21.387695+00:00"},{"alias_kind":"pith_short_16","alias_value":"LDOSFEOGVFXXJI5H","created_at":"2026-05-18T12:33:21.387695+00:00"},{"alias_kind":"pith_short_8","alias_value":"LDOSFEOG","created_at":"2026-05-18T12:33:21.387695+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.06992","citing_title":"Why Does Agentic Safety Fail to Generalize Across Tasks?","ref_index":55,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC","json":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC.json","graph_json":"https://pith.science/api/pith-number/LDOSFEOGVFXXJI5HS57ZRTOYXC/graph.json","events_json":"https://pith.science/api/pith-number/LDOSFEOGVFXXJI5HS57ZRTOYXC/events.json","paper":"https://pith.science/paper/LDOSFEOG"},"agent_actions":{"view_html":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC","download_json":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC.json","view_paper":"https://pith.science/paper/LDOSFEOG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1907.01475&json=true","fetch_graph":"https://pith.science/api/pith-number/LDOSFEOGVFXXJI5HS57ZRTOYXC/graph.json","fetch_events":"https://pith.science/api/pith-number/LDOSFEOGVFXXJI5HS57ZRTOYXC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC/action/storage_attestation","attest_author":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC/action/author_attestation","sign_citation":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC/action/citation_signature","submit_replication":"https://pith.science/pith/LDOSFEOGVFXXJI5HS57ZRTOYXC/action/replication_record"}},"created_at":"2026-05-17T23:41:40.297395+00:00","updated_at":"2026-05-17T23:41:40.297395+00:00"}