{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:T4MMAUXIQFSLZRUXZKKIG7K6YR","short_pith_number":"pith:T4MMAUXI","schema_version":"1.0","canonical_sha256":"9f18c052e88164bcc697ca94837d5ec457accd2b7a854a19a199d254de75bdb4","source":{"kind":"arxiv","id":"2607.08925","version":1},"attestation_state":"computed","paper":{"title":"SafeExplorer: An Unbiased Policy Gradient for Reinforcement Learning with Recovery Interventions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Elham Daneshmand, Glen Berseth, Hsiu-Chin Lin, Majid Khadiv","submitted_at":"2026-07-09T20:41:45Z","abstract_excerpt":"Training reinforcement-learning agents directly on physical robots makes every fall costly, since a fall can damage the platform and cannot be undone like a simulator reset; the goal is therefore to minimize falls during training rather than trade them off against return, as constrained Markov decision process (MDP) formulations do. A standard mitigation hands control to a separate recovery policy whenever the agent leaves a designer-specified safe region (a subset of state space it should stay within), but the resulting mixed-policy rollouts silently bias every on-policy update, and the impor"},"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":"2607.08925","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-09T20:41:45Z","cross_cats_sorted":[],"title_canon_sha256":"d69c037e15e260b13993dd23d0a9d4e29de85d6aede1a04cbbcece6b1cc5fbad","abstract_canon_sha256":"c1a0c0f1b7ed8123312f7826bb338297534e748786407f955eef44f9bbe351d7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-13T00:17:27.359965Z","signature_b64":"XKq9i8r2OCN7Rv7LjiuLcsBcg2aBeVizcBCxlqq2J8vzFC5s8IyOVKVtWZSCpYZYBLSWGW/cRySGXphSrtkBCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9f18c052e88164bcc697ca94837d5ec457accd2b7a854a19a199d254de75bdb4","last_reissued_at":"2026-07-13T00:17:27.358969Z","signature_status":"signed_v1","first_computed_at":"2026-07-13T00:17:27.358969Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeExplorer: An Unbiased Policy Gradient for Reinforcement Learning with Recovery Interventions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Elham Daneshmand, Glen Berseth, Hsiu-Chin Lin, Majid Khadiv","submitted_at":"2026-07-09T20:41:45Z","abstract_excerpt":"Training reinforcement-learning agents directly on physical robots makes every fall costly, since a fall can damage the platform and cannot be undone like a simulator reset; the goal is therefore to minimize falls during training rather than trade them off against return, as constrained Markov decision process (MDP) formulations do. A standard mitigation hands control to a separate recovery policy whenever the agent leaves a designer-specified safe region (a subset of state space it should stay within), but the resulting mixed-policy rollouts silently bias every on-policy update, and the impor"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.08925","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/2607.08925/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":"2607.08925","created_at":"2026-07-13T00:17:27.359491+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.08925v1","created_at":"2026-07-13T00:17:27.359491+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.08925","created_at":"2026-07-13T00:17:27.359491+00:00"},{"alias_kind":"pith_short_12","alias_value":"T4MMAUXIQFSL","created_at":"2026-07-13T00:17:27.359491+00:00"},{"alias_kind":"pith_short_16","alias_value":"T4MMAUXIQFSLZRUX","created_at":"2026-07-13T00:17:27.359491+00:00"},{"alias_kind":"pith_short_8","alias_value":"T4MMAUXI","created_at":"2026-07-13T00:17:27.359491+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR","json":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR.json","graph_json":"https://pith.science/api/pith-number/T4MMAUXIQFSLZRUXZKKIG7K6YR/graph.json","events_json":"https://pith.science/api/pith-number/T4MMAUXIQFSLZRUXZKKIG7K6YR/events.json","paper":"https://pith.science/paper/T4MMAUXI"},"agent_actions":{"view_html":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR","download_json":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR.json","view_paper":"https://pith.science/paper/T4MMAUXI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.08925&json=true","fetch_graph":"https://pith.science/api/pith-number/T4MMAUXIQFSLZRUXZKKIG7K6YR/graph.json","fetch_events":"https://pith.science/api/pith-number/T4MMAUXIQFSLZRUXZKKIG7K6YR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR/action/storage_attestation","attest_author":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR/action/author_attestation","sign_citation":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR/action/citation_signature","submit_replication":"https://pith.science/pith/T4MMAUXIQFSLZRUXZKKIG7K6YR/action/replication_record"}},"created_at":"2026-07-13T00:17:27.359491+00:00","updated_at":"2026-07-13T00:17:27.359491+00:00"}