{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:HWMCJR4P4MMUKCOC3FGRGPTJLY","short_pith_number":"pith:HWMCJR4P","schema_version":"1.0","canonical_sha256":"3d9824c78fe3194509c2d94d133e695e10859a73782a7c6e565fd0e380604ba6","source":{"kind":"arxiv","id":"1811.02553","version":4},"attestation_state":"computed","paper":{"title":"A Closer Look at Deep Policy Gradients","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.NE","cs.RO","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aleksander Madry, Andrew Ilyas, Dimitris Tsipras, Firdaus Janoos, Larry Rudolph, Logan Engstrom, Shibani Santurkar","submitted_at":"2018-11-06T18:54:21Z","abstract_excerpt":"We study how the behavior of deep policy gradient algorithms reflects the conceptual framework motivating their development. To this end, we propose a fine-grained analysis of state-of-the-art methods based on key elements of this framework: gradient estimation, value prediction, and optimization landscapes. Our results show that the behavior of deep policy gradient algorithms often deviates from what their motivating framework would predict: the surrogate objective does not match the true reward landscape, learned value estimators fail to fit the true value function, and gradient estimates po"},"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":"1811.02553","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2018-11-06T18:54:21Z","cross_cats_sorted":["cs.NE","cs.RO","stat.ML"],"title_canon_sha256":"5d7ae50a174e750b5777e4903fddfe8f795b812c5354238e18b15fa0865ef164","abstract_canon_sha256":"515f17763edab563545009f7f42ca159276ae4126ad932244f376b29438cd760"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:05:22.860250Z","signature_b64":"sfgTyayLhY7wZhEGh4PkPKVuoqV8dPs4W2gaAes1V3w+RAZB5h/vE7HX5r7XD58+JSxGNaVJwZmRqPgBvpPfCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3d9824c78fe3194509c2d94d133e695e10859a73782a7c6e565fd0e380604ba6","last_reissued_at":"2026-07-05T01:05:22.859768Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:05:22.859768Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Closer Look at Deep Policy Gradients","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.NE","cs.RO","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aleksander Madry, Andrew Ilyas, Dimitris Tsipras, Firdaus Janoos, Larry Rudolph, Logan Engstrom, Shibani Santurkar","submitted_at":"2018-11-06T18:54:21Z","abstract_excerpt":"We study how the behavior of deep policy gradient algorithms reflects the conceptual framework motivating their development. To this end, we propose a fine-grained analysis of state-of-the-art methods based on key elements of this framework: gradient estimation, value prediction, and optimization landscapes. Our results show that the behavior of deep policy gradient algorithms often deviates from what their motivating framework would predict: the surrogate objective does not match the true reward landscape, learned value estimators fail to fit the true value function, and gradient estimates po"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1811.02553","kind":"arxiv","version":4},"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/1811.02553/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":"1811.02553","created_at":"2026-07-05T01:05:22.859827+00:00"},{"alias_kind":"arxiv_version","alias_value":"1811.02553v4","created_at":"2026-07-05T01:05:22.859827+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1811.02553","created_at":"2026-07-05T01:05:22.859827+00:00"},{"alias_kind":"pith_short_12","alias_value":"HWMCJR4P4MMU","created_at":"2026-07-05T01:05:22.859827+00:00"},{"alias_kind":"pith_short_16","alias_value":"HWMCJR4P4MMUKCOC","created_at":"2026-07-05T01:05:22.859827+00:00"},{"alias_kind":"pith_short_8","alias_value":"HWMCJR4P","created_at":"2026-07-05T01:05:22.859827+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2102.01293","citing_title":"Scaling Laws for Transfer","ref_index":105,"is_internal_anchor":false},{"citing_arxiv_id":"2112.00861","citing_title":"A General Language Assistant as a Laboratory for Alignment","ref_index":147,"is_internal_anchor":false},{"citing_arxiv_id":"2207.05221","citing_title":"Language Models (Mostly) Know What They Know","ref_index":224,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02320","citing_title":"ANO: A Principled Approach to Robust Policy Optimization","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY","json":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY.json","graph_json":"https://pith.science/api/pith-number/HWMCJR4P4MMUKCOC3FGRGPTJLY/graph.json","events_json":"https://pith.science/api/pith-number/HWMCJR4P4MMUKCOC3FGRGPTJLY/events.json","paper":"https://pith.science/paper/HWMCJR4P"},"agent_actions":{"view_html":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY","download_json":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY.json","view_paper":"https://pith.science/paper/HWMCJR4P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1811.02553&json=true","fetch_graph":"https://pith.science/api/pith-number/HWMCJR4P4MMUKCOC3FGRGPTJLY/graph.json","fetch_events":"https://pith.science/api/pith-number/HWMCJR4P4MMUKCOC3FGRGPTJLY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY/action/storage_attestation","attest_author":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY/action/author_attestation","sign_citation":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY/action/citation_signature","submit_replication":"https://pith.science/pith/HWMCJR4P4MMUKCOC3FGRGPTJLY/action/replication_record"}},"created_at":"2026-07-05T01:05:22.859827+00:00","updated_at":"2026-07-05T01:05:22.859827+00:00"}