{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XZXTEEFOIN7QN3KAQJYDSPJTO6","short_pith_number":"pith:XZXTEEFO","schema_version":"1.0","canonical_sha256":"be6f3210ae437f06ed408270393d3377be4af80d9ec4a252e27b918e5cd8f797","source":{"kind":"arxiv","id":"2405.18180","version":3},"attestation_state":"computed","paper":{"title":"Safe Reinforcement Learning in Black-Box Environments via Adaptive Shielding","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Calum Imrie, Daniel Bethell, Radu Calinescu, Simos Gerasimou","submitted_at":"2024-05-28T13:47:21Z","abstract_excerpt":"Empowering safe exploration of reinforcement learning (RL) agents during training is a critical challenge towards their deployment in many real-world scenarios. When prior knowledge of the domain or task is unavailable, training RL agents in unknown, black-box environments presents an even greater safety risk. We introduce ADVICE (Adaptive Shielding with a Contrastive Autoencoder), a novel post-shielding technique that distinguishes safe and unsafe features of state-action pairs during training, and uses this knowledge to protect the RL agent from executing actions that yield likely hazardous "},"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":"2405.18180","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.AI","submitted_at":"2024-05-28T13:47:21Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"ecbd762d44f9068e214d80a65ed2d0b42461538a15d7720e530428c93e663c92","abstract_canon_sha256":"7e384665bb00fb7d8269822d1ea7509ee0d79d21a9172ec261424d392c70b0ed"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:59:05.376347Z","signature_b64":"sl/Wd0Nga1HD412vNjtBa6dCRmROIjFnigxpCqWgWE9maG9/qHc1wqlpECdNVtYht5vlpE9PfQFeY60dN6nlBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be6f3210ae437f06ed408270393d3377be4af80d9ec4a252e27b918e5cd8f797","last_reissued_at":"2026-07-05T11:59:05.375834Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:59:05.375834Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Safe Reinforcement Learning in Black-Box Environments via Adaptive Shielding","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Calum Imrie, Daniel Bethell, Radu Calinescu, Simos Gerasimou","submitted_at":"2024-05-28T13:47:21Z","abstract_excerpt":"Empowering safe exploration of reinforcement learning (RL) agents during training is a critical challenge towards their deployment in many real-world scenarios. When prior knowledge of the domain or task is unavailable, training RL agents in unknown, black-box environments presents an even greater safety risk. We introduce ADVICE (Adaptive Shielding with a Contrastive Autoencoder), a novel post-shielding technique that distinguishes safe and unsafe features of state-action pairs during training, and uses this knowledge to protect the RL agent from executing actions that yield likely hazardous "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.18180","kind":"arxiv","version":3},"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/2405.18180/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":"2405.18180","created_at":"2026-07-05T11:59:05.375897+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.18180v3","created_at":"2026-07-05T11:59:05.375897+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.18180","created_at":"2026-07-05T11:59:05.375897+00:00"},{"alias_kind":"pith_short_12","alias_value":"XZXTEEFOIN7Q","created_at":"2026-07-05T11:59:05.375897+00:00"},{"alias_kind":"pith_short_16","alias_value":"XZXTEEFOIN7QN3KA","created_at":"2026-07-05T11:59:05.375897+00:00"},{"alias_kind":"pith_short_8","alias_value":"XZXTEEFO","created_at":"2026-07-05T11:59:05.375897+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.30576","citing_title":"Uncertainty-Aware and Temporally Regulated Expert Advice in Reinforcement Learning for Autonomous Driving","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6","json":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6.json","graph_json":"https://pith.science/api/pith-number/XZXTEEFOIN7QN3KAQJYDSPJTO6/graph.json","events_json":"https://pith.science/api/pith-number/XZXTEEFOIN7QN3KAQJYDSPJTO6/events.json","paper":"https://pith.science/paper/XZXTEEFO"},"agent_actions":{"view_html":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6","download_json":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6.json","view_paper":"https://pith.science/paper/XZXTEEFO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.18180&json=true","fetch_graph":"https://pith.science/api/pith-number/XZXTEEFOIN7QN3KAQJYDSPJTO6/graph.json","fetch_events":"https://pith.science/api/pith-number/XZXTEEFOIN7QN3KAQJYDSPJTO6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6/action/storage_attestation","attest_author":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6/action/author_attestation","sign_citation":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6/action/citation_signature","submit_replication":"https://pith.science/pith/XZXTEEFOIN7QN3KAQJYDSPJTO6/action/replication_record"}},"created_at":"2026-07-05T11:59:05.375897+00:00","updated_at":"2026-07-05T11:59:05.375897+00:00"}