{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:MBXQASGDISW4ZDSFKKR45WO3NN","short_pith_number":"pith:MBXQASGD","schema_version":"1.0","canonical_sha256":"606f0048c344adcc8e4552a3ced9db6b78d6cd700fb2e96b1f1acec5a3c75ac5","source":{"kind":"arxiv","id":"2005.12729","version":1},"attestation_state":"computed","paper":{"title":"Implementation Matters in Deep Policy Gradients: A Case Study on PPO and TRPO","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["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":"2020-05-25T16:24:59Z","abstract_excerpt":"We study the roots of algorithmic progress in deep policy gradient algorithms through a case study on two popular algorithms: Proximal Policy Optimization (PPO) and Trust Region Policy Optimization (TRPO). Specifically, we investigate the consequences of \"code-level optimizations:\" algorithm augmentations found only in implementations or described as auxiliary details to the core algorithm. Seemingly of secondary importance, such optimizations turn out to have a major impact on agent behavior. Our results show that they (a) are responsible for most of PPO's gain in cumulative reward over TRPO,"},"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":"2005.12729","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-05-25T16:24:59Z","cross_cats_sorted":["cs.RO","stat.ML"],"title_canon_sha256":"cb1f00c1e5dc8378790e6426e735fe837141644fa5b1db39c8f848bc14a13bc5","abstract_canon_sha256":"edeedce99e127960f674cd8ab615e6fbafc8b0dad5ef447fcff08a69c16e99d0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:05:56.921157Z","signature_b64":"RdLi93YgU313kUEBGx4X+N0RPNVUnWZjmAHSv/w8cbT2dUaV79iz5vhANMq8N8juuAE/EJ1pUkNvQIMQaBxQCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"606f0048c344adcc8e4552a3ced9db6b78d6cd700fb2e96b1f1acec5a3c75ac5","last_reissued_at":"2026-07-05T01:05:56.920631Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:05:56.920631Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Implementation Matters in Deep Policy Gradients: A Case Study on PPO and TRPO","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["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":"2020-05-25T16:24:59Z","abstract_excerpt":"We study the roots of algorithmic progress in deep policy gradient algorithms through a case study on two popular algorithms: Proximal Policy Optimization (PPO) and Trust Region Policy Optimization (TRPO). Specifically, we investigate the consequences of \"code-level optimizations:\" algorithm augmentations found only in implementations or described as auxiliary details to the core algorithm. Seemingly of secondary importance, such optimizations turn out to have a major impact on agent behavior. Our results show that they (a) are responsible for most of PPO's gain in cumulative reward over TRPO,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2005.12729","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/2005.12729/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":"2005.12729","created_at":"2026-07-05T01:05:56.920691+00:00"},{"alias_kind":"arxiv_version","alias_value":"2005.12729v1","created_at":"2026-07-05T01:05:56.920691+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2005.12729","created_at":"2026-07-05T01:05:56.920691+00:00"},{"alias_kind":"pith_short_12","alias_value":"MBXQASGDISW4","created_at":"2026-07-05T01:05:56.920691+00:00"},{"alias_kind":"pith_short_16","alias_value":"MBXQASGDISW4ZDSF","created_at":"2026-07-05T01:05:56.920691+00:00"},{"alias_kind":"pith_short_8","alias_value":"MBXQASGD","created_at":"2026-07-05T01:05:56.920691+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.23110","citing_title":"LOLLA: Deep Reinforcement Learning for Closed-Loop Link Adaptation Towards a GPU-Accelerated AI-RAN","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2606.13605","citing_title":"Distribution-Agnostic Robust Trajectory Optimization via Chance-Constrained Reinforcement Learning","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04574","citing_title":"Dynamic Multi-Pair Trading Strategy in Cryptocurrency Markets with Deep Reinforcement Learning","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26784","citing_title":"Ratio-Variance Regularized Policy Optimization","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2408.07199","citing_title":"Agent Q: Advanced Reasoning and Learning for Autonomous AI Agents","ref_index":208,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26126","citing_title":"Application of Deep Reinforcement Learning to Event-Triggered Control for Networked Artificial Pancreas Systems","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2304.06767","citing_title":"RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment","ref_index":94,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11473","citing_title":"TOPPO: Rethinking PPO for Multi-Task Reinforcement Learning with Critic Balancing","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26126","citing_title":"Application of Deep Reinforcement Learning to Event-Triggered Control for Networked Artificial Pancreas Systems","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06032","citing_title":"Does Synthetic Data Help? Empirical Evidence from Deep Learning Time Series Forecasters","ref_index":254,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06987","citing_title":"Response Time Enhances Alignment with Heterogeneous Preferences","ref_index":66,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18578","citing_title":"Bounded Ratio Reinforcement Learning","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02320","citing_title":"ANO: A Principled Approach to Robust Policy Optimization","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN","json":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN.json","graph_json":"https://pith.science/api/pith-number/MBXQASGDISW4ZDSFKKR45WO3NN/graph.json","events_json":"https://pith.science/api/pith-number/MBXQASGDISW4ZDSFKKR45WO3NN/events.json","paper":"https://pith.science/paper/MBXQASGD"},"agent_actions":{"view_html":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN","download_json":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN.json","view_paper":"https://pith.science/paper/MBXQASGD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2005.12729&json=true","fetch_graph":"https://pith.science/api/pith-number/MBXQASGDISW4ZDSFKKR45WO3NN/graph.json","fetch_events":"https://pith.science/api/pith-number/MBXQASGDISW4ZDSFKKR45WO3NN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN/action/storage_attestation","attest_author":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN/action/author_attestation","sign_citation":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN/action/citation_signature","submit_replication":"https://pith.science/pith/MBXQASGDISW4ZDSFKKR45WO3NN/action/replication_record"}},"created_at":"2026-07-05T01:05:56.920691+00:00","updated_at":"2026-07-05T01:05:56.920691+00:00"}