{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:BGFYATXLUFCO4XEOIJYQSHWSTO","short_pith_number":"pith:BGFYATXL","schema_version":"1.0","canonical_sha256":"098b804eeba144ee5c8e4271091ed29b8a0a1953697316f263ebb53d3877f0a0","source":{"kind":"arxiv","id":"2107.06277","version":1},"attestation_state":"computed","paper":{"title":"Why Generalization in RL is Difficult: Epistemic POMDPs and Implicit Partial Observability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Amy Zhang, Aviral Kumar, Dibya Ghosh, Jad Rahme, Ryan P. Adams, Sergey Levine","submitted_at":"2021-07-13T17:59:25Z","abstract_excerpt":"Generalization is a central challenge for the deployment of reinforcement learning (RL) systems in the real world. In this paper, we show that the sequential structure of the RL problem necessitates new approaches to generalization beyond the well-studied techniques used in supervised learning. While supervised learning methods can generalize effectively without explicitly accounting for epistemic uncertainty, we show that, perhaps surprisingly, this is not the case in RL. We show that generalization to unseen test conditions from a limited number of training conditions induces implicit partia"},"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":"2107.06277","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-07-13T17:59:25Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"98919c59dc84d2c537b3ef077ca9af035bc3fd581effb6b375e29b5bfa3f7fbf","abstract_canon_sha256":"563b73126bca10b64a047dc34ea667d2dcfccab5dcaeacd0032b6f578582c2ae"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:57:32.466746Z","signature_b64":"S6SLeA1sH0rXOXMtQC2Mst7fmpA1eVU+lJgCm443QG4w4jz77J+khj9t3Dj3ypGHun0suDYO27gf6eEqxQhIBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"098b804eeba144ee5c8e4271091ed29b8a0a1953697316f263ebb53d3877f0a0","last_reissued_at":"2026-07-05T02:57:32.466157Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:57:32.466157Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Why Generalization in RL is Difficult: Epistemic POMDPs and Implicit Partial Observability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Amy Zhang, Aviral Kumar, Dibya Ghosh, Jad Rahme, Ryan P. Adams, Sergey Levine","submitted_at":"2021-07-13T17:59:25Z","abstract_excerpt":"Generalization is a central challenge for the deployment of reinforcement learning (RL) systems in the real world. In this paper, we show that the sequential structure of the RL problem necessitates new approaches to generalization beyond the well-studied techniques used in supervised learning. While supervised learning methods can generalize effectively without explicitly accounting for epistemic uncertainty, we show that, perhaps surprisingly, this is not the case in RL. We show that generalization to unseen test conditions from a limited number of training conditions induces implicit partia"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2107.06277","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/2107.06277/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":"2107.06277","created_at":"2026-07-05T02:57:32.466232+00:00"},{"alias_kind":"arxiv_version","alias_value":"2107.06277v1","created_at":"2026-07-05T02:57:32.466232+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2107.06277","created_at":"2026-07-05T02:57:32.466232+00:00"},{"alias_kind":"pith_short_12","alias_value":"BGFYATXLUFCO","created_at":"2026-07-05T02:57:32.466232+00:00"},{"alias_kind":"pith_short_16","alias_value":"BGFYATXLUFCO4XEO","created_at":"2026-07-05T02:57:32.466232+00:00"},{"alias_kind":"pith_short_8","alias_value":"BGFYATXL","created_at":"2026-07-05T02:57:32.466232+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.09671","citing_title":"Belief-State RWKV for Reinforcement Learning under Partial Observability","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO","json":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO.json","graph_json":"https://pith.science/api/pith-number/BGFYATXLUFCO4XEOIJYQSHWSTO/graph.json","events_json":"https://pith.science/api/pith-number/BGFYATXLUFCO4XEOIJYQSHWSTO/events.json","paper":"https://pith.science/paper/BGFYATXL"},"agent_actions":{"view_html":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO","download_json":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO.json","view_paper":"https://pith.science/paper/BGFYATXL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2107.06277&json=true","fetch_graph":"https://pith.science/api/pith-number/BGFYATXLUFCO4XEOIJYQSHWSTO/graph.json","fetch_events":"https://pith.science/api/pith-number/BGFYATXLUFCO4XEOIJYQSHWSTO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO/action/storage_attestation","attest_author":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO/action/author_attestation","sign_citation":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO/action/citation_signature","submit_replication":"https://pith.science/pith/BGFYATXLUFCO4XEOIJYQSHWSTO/action/replication_record"}},"created_at":"2026-07-05T02:57:32.466232+00:00","updated_at":"2026-07-05T02:57:32.466232+00:00"}