{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:G7YAM22NQP2TGCQ3ZNTCWSHLRK","short_pith_number":"pith:G7YAM22N","schema_version":"1.0","canonical_sha256":"37f0066b4d83f5330a1bcb662b48eb8ab518c01e0dbe2a6047dabce42e4a053f","source":{"kind":"arxiv","id":"1904.11455","version":1},"attestation_state":"computed","paper":{"title":"Ray Interference: a Source of Plateaus in Deep Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Diana Borsa, Joseph Modayil, Razvan Pascanu, Tom Schaul","submitted_at":"2019-04-25T16:54:02Z","abstract_excerpt":"Rather than proposing a new method, this paper investigates an issue present in existing learning algorithms. We study the learning dynamics of reinforcement learning (RL), specifically a characteristic coupling between learning and data generation that arises because RL agents control their future data distribution. In the presence of function approximation, this coupling can lead to a problematic type of 'ray interference', characterized by learning dynamics that sequentially traverse a number of performance plateaus, effectively constraining the agent to learn one thing at a time even when "},"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":"1904.11455","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-04-25T16:54:02Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"52d4e8815f5ce59f1e6f7614c18233052363d02b154ed3f68100a1114c6025d2","abstract_canon_sha256":"4377158266573881816c8428faa75b3f3c296710788e971c7b02b63e3f887226"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:47:44.818536Z","signature_b64":"LNN6+GY9Mi6+wchnmeIOiEh1UvVCZqKoPQQQIJ7iTqfh01CQ1qE3YRhujT7VH+t/7caMzh6ddnmmWdlbvE2xDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"37f0066b4d83f5330a1bcb662b48eb8ab518c01e0dbe2a6047dabce42e4a053f","last_reissued_at":"2026-05-17T23:47:44.817956Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:47:44.817956Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Ray Interference: a Source of Plateaus in Deep Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Diana Borsa, Joseph Modayil, Razvan Pascanu, Tom Schaul","submitted_at":"2019-04-25T16:54:02Z","abstract_excerpt":"Rather than proposing a new method, this paper investigates an issue present in existing learning algorithms. We study the learning dynamics of reinforcement learning (RL), specifically a characteristic coupling between learning and data generation that arises because RL agents control their future data distribution. In the presence of function approximation, this coupling can lead to a problematic type of 'ray interference', characterized by learning dynamics that sequentially traverse a number of performance plateaus, effectively constraining the agent to learn one thing at a time even when "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1904.11455","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":""},"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":"1904.11455","created_at":"2026-05-17T23:47:44.818032+00:00"},{"alias_kind":"arxiv_version","alias_value":"1904.11455v1","created_at":"2026-05-17T23:47:44.818032+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1904.11455","created_at":"2026-05-17T23:47:44.818032+00:00"},{"alias_kind":"pith_short_12","alias_value":"G7YAM22NQP2T","created_at":"2026-05-18T12:33:18.533446+00:00"},{"alias_kind":"pith_short_16","alias_value":"G7YAM22NQP2TGCQ3","created_at":"2026-05-18T12:33:18.533446+00:00"},{"alias_kind":"pith_short_8","alias_value":"G7YAM22N","created_at":"2026-05-18T12:33:18.533446+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"1907.02057","citing_title":"Benchmarking Model-Based Reinforcement Learning","ref_index":42,"is_internal_anchor":true},{"citing_arxiv_id":"2309.16797","citing_title":"Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution","ref_index":288,"is_internal_anchor":true},{"citing_arxiv_id":"2605.03363","citing_title":"Learning Reactive Dexterous Grasping via Hierarchical Task-Space RL Planning and Joint-Space QP Control","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10909","citing_title":"Revisiting Policy Gradients for Restricted Policy Classes: Escaping Myopic Local Optima with $k$-step Policy Gradients","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25872","citing_title":"When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy Gradient","ref_index":73,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK","json":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK.json","graph_json":"https://pith.science/api/pith-number/G7YAM22NQP2TGCQ3ZNTCWSHLRK/graph.json","events_json":"https://pith.science/api/pith-number/G7YAM22NQP2TGCQ3ZNTCWSHLRK/events.json","paper":"https://pith.science/paper/G7YAM22N"},"agent_actions":{"view_html":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK","download_json":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK.json","view_paper":"https://pith.science/paper/G7YAM22N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1904.11455&json=true","fetch_graph":"https://pith.science/api/pith-number/G7YAM22NQP2TGCQ3ZNTCWSHLRK/graph.json","fetch_events":"https://pith.science/api/pith-number/G7YAM22NQP2TGCQ3ZNTCWSHLRK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK/action/storage_attestation","attest_author":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK/action/author_attestation","sign_citation":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK/action/citation_signature","submit_replication":"https://pith.science/pith/G7YAM22NQP2TGCQ3ZNTCWSHLRK/action/replication_record"}},"created_at":"2026-05-17T23:47:44.818032+00:00","updated_at":"2026-05-17T23:47:44.818032+00:00"}