{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:XQMZ2XJZBU7CTUBY2YSA25UM2H","short_pith_number":"pith:XQMZ2XJZ","schema_version":"1.0","canonical_sha256":"bc199d5d390d3e29d038d6240d768cd1fe589fe7081ce7c2ae656244ec46d03f","source":{"kind":"arxiv","id":"2106.06630","version":1},"attestation_state":"computed","paper":{"title":"Corruption-Robust Offline Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jerry Zhu, Wen Sun, Xuezhou Zhang, Yiding Chen","submitted_at":"2021-06-11T22:41:53Z","abstract_excerpt":"We study the adversarial robustness in offline reinforcement learning. Given a batch dataset consisting of tuples $(s, a, r, s')$, an adversary is allowed to arbitrarily modify $\\epsilon$ fraction of the tuples. From the corrupted dataset the learner aims to robustly identify a near-optimal policy. We first show that a worst-case $\\Omega(d\\epsilon)$ optimality gap is unavoidable in linear MDP of dimension $d$, even if the adversary only corrupts the reward element in a tuple. This contrasts with dimension-free results in robust supervised learning and best-known lower-bound in the online RL se"},"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":"2106.06630","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2021-06-11T22:41:53Z","cross_cats_sorted":[],"title_canon_sha256":"ae9dc094ff4f2cfb3ebd3ec5948ed1ea535e863375da5a29b0969494baf6a0bc","abstract_canon_sha256":"453c1546a8a2920109a542ad674be591b4c5f592f8e5fda152b3781e452186e4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:48:39.754597Z","signature_b64":"AA2jBRK55xGNgcWGC9z/tfKKWLMmxrOIvYvie8BJZDO9ypbGzETey/VMj0333N6ix6W1mp+1wDknPkD2J8SrAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc199d5d390d3e29d038d6240d768cd1fe589fe7081ce7c2ae656244ec46d03f","last_reissued_at":"2026-07-05T02:48:39.754120Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:48:39.754120Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Corruption-Robust Offline Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jerry Zhu, Wen Sun, Xuezhou Zhang, Yiding Chen","submitted_at":"2021-06-11T22:41:53Z","abstract_excerpt":"We study the adversarial robustness in offline reinforcement learning. Given a batch dataset consisting of tuples $(s, a, r, s')$, an adversary is allowed to arbitrarily modify $\\epsilon$ fraction of the tuples. From the corrupted dataset the learner aims to robustly identify a near-optimal policy. We first show that a worst-case $\\Omega(d\\epsilon)$ optimality gap is unavoidable in linear MDP of dimension $d$, even if the adversary only corrupts the reward element in a tuple. This contrasts with dimension-free results in robust supervised learning and best-known lower-bound in the online RL se"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.06630","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/2106.06630/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":"2106.06630","created_at":"2026-07-05T02:48:39.754178+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.06630v1","created_at":"2026-07-05T02:48:39.754178+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.06630","created_at":"2026-07-05T02:48:39.754178+00:00"},{"alias_kind":"pith_short_12","alias_value":"XQMZ2XJZBU7C","created_at":"2026-07-05T02:48:39.754178+00:00"},{"alias_kind":"pith_short_16","alias_value":"XQMZ2XJZBU7CTUBY","created_at":"2026-07-05T02:48:39.754178+00:00"},{"alias_kind":"pith_short_8","alias_value":"XQMZ2XJZ","created_at":"2026-07-05T02:48:39.754178+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22579","citing_title":"Stationary Robust Mean-Field Games under Model Mismatches","ref_index":116,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H","json":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H.json","graph_json":"https://pith.science/api/pith-number/XQMZ2XJZBU7CTUBY2YSA25UM2H/graph.json","events_json":"https://pith.science/api/pith-number/XQMZ2XJZBU7CTUBY2YSA25UM2H/events.json","paper":"https://pith.science/paper/XQMZ2XJZ"},"agent_actions":{"view_html":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H","download_json":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H.json","view_paper":"https://pith.science/paper/XQMZ2XJZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.06630&json=true","fetch_graph":"https://pith.science/api/pith-number/XQMZ2XJZBU7CTUBY2YSA25UM2H/graph.json","fetch_events":"https://pith.science/api/pith-number/XQMZ2XJZBU7CTUBY2YSA25UM2H/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H/action/storage_attestation","attest_author":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H/action/author_attestation","sign_citation":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H/action/citation_signature","submit_replication":"https://pith.science/pith/XQMZ2XJZBU7CTUBY2YSA25UM2H/action/replication_record"}},"created_at":"2026-07-05T02:48:39.754178+00:00","updated_at":"2026-07-05T02:48:39.754178+00:00"}