{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:CB64UDTSHVN2HB6FSYMBVSUN6Q","short_pith_number":"pith:CB64UDTS","schema_version":"1.0","canonical_sha256":"107dca0e723d5ba387c596181aca8df43ee220986d8437bfd7001c83db8447bf","source":{"kind":"arxiv","id":"2010.14603","version":1},"attestation_state":"computed","paper":{"title":"Learning to be Safe: Deep RL with a Safety Critic","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.RO"],"primary_cat":"cs.LG","authors_text":"Benjamin Eysenbach, Chelsea Finn, Jie Tan, Krishnan Srinivasan, Sehoon Ha","submitted_at":"2020-10-27T20:53:20Z","abstract_excerpt":"Safety is an essential component for deploying reinforcement learning (RL) algorithms in real-world scenarios, and is critical during the learning process itself. A natural first approach toward safe RL is to manually specify constraints on the policy's behavior. However, just as learning has enabled progress in large-scale development of AI systems, learning safety specifications may also be necessary to ensure safety in messy open-world environments where manual safety specifications cannot scale. Akin to how humans learn incrementally starting in child-safe environments, we propose to learn"},"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":"2010.14603","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-27T20:53:20Z","cross_cats_sorted":["cs.RO"],"title_canon_sha256":"4259f9175dc4a8b8fee05d1c737b12d98341d06a1e25d50c3151343c1d4c9309","abstract_canon_sha256":"d59d69804485cb41c1a6fa8f65c9dc67037bba4b38c167e2714ce1ba947dfd99"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:47:16.813429Z","signature_b64":"C9cTWcX7uZTkqcqoojKaGjCfrgNoNDD3wBFxas0v0vp6uJ2tyJ95nbsNmr1mmk8FYKl8dBskJQ0hnpF5YG71Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"107dca0e723d5ba387c596181aca8df43ee220986d8437bfd7001c83db8447bf","last_reissued_at":"2026-07-05T01:47:16.812960Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:47:16.812960Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning to be Safe: Deep RL with a Safety Critic","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.RO"],"primary_cat":"cs.LG","authors_text":"Benjamin Eysenbach, Chelsea Finn, Jie Tan, Krishnan Srinivasan, Sehoon Ha","submitted_at":"2020-10-27T20:53:20Z","abstract_excerpt":"Safety is an essential component for deploying reinforcement learning (RL) algorithms in real-world scenarios, and is critical during the learning process itself. A natural first approach toward safe RL is to manually specify constraints on the policy's behavior. However, just as learning has enabled progress in large-scale development of AI systems, learning safety specifications may also be necessary to ensure safety in messy open-world environments where manual safety specifications cannot scale. Akin to how humans learn incrementally starting in child-safe environments, we propose to learn"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.14603","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/2010.14603/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":"2010.14603","created_at":"2026-07-05T01:47:16.813018+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.14603v1","created_at":"2026-07-05T01:47:16.813018+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.14603","created_at":"2026-07-05T01:47:16.813018+00:00"},{"alias_kind":"pith_short_12","alias_value":"CB64UDTSHVN2","created_at":"2026-07-05T01:47:16.813018+00:00"},{"alias_kind":"pith_short_16","alias_value":"CB64UDTSHVN2HB6F","created_at":"2026-07-05T01:47:16.813018+00:00"},{"alias_kind":"pith_short_8","alias_value":"CB64UDTS","created_at":"2026-07-05T01:47:16.813018+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.12372","citing_title":"UniIntervene: Agentic Intervention for Efficient Real-World Reinforcement Learning","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10228","citing_title":"SHAPO: Sharpness-Aware Policy Optimization for Safe Exploration","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2402.05284","citing_title":"Analyzing Adversarial Inputs in Deep Reinforcement Learning","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2408.08812","citing_title":"TRAM: Test-Time Risk Adaptation with Mixture of Agents","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19737","citing_title":"Safe Continual Reinforcement Learning in Non-stationary Environments","ref_index":36,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q","json":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q.json","graph_json":"https://pith.science/api/pith-number/CB64UDTSHVN2HB6FSYMBVSUN6Q/graph.json","events_json":"https://pith.science/api/pith-number/CB64UDTSHVN2HB6FSYMBVSUN6Q/events.json","paper":"https://pith.science/paper/CB64UDTS"},"agent_actions":{"view_html":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q","download_json":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q.json","view_paper":"https://pith.science/paper/CB64UDTS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.14603&json=true","fetch_graph":"https://pith.science/api/pith-number/CB64UDTSHVN2HB6FSYMBVSUN6Q/graph.json","fetch_events":"https://pith.science/api/pith-number/CB64UDTSHVN2HB6FSYMBVSUN6Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q/action/storage_attestation","attest_author":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q/action/author_attestation","sign_citation":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q/action/citation_signature","submit_replication":"https://pith.science/pith/CB64UDTSHVN2HB6FSYMBVSUN6Q/action/replication_record"}},"created_at":"2026-07-05T01:47:16.813018+00:00","updated_at":"2026-07-05T01:47:16.813018+00:00"}