{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:WID7QOXWKJQE35BICP3WHMYKGE","short_pith_number":"pith:WID7QOXW","schema_version":"1.0","canonical_sha256":"b207f83af652604df42813f763b30a3125ee0b196676c3ead9123dfe78b4f6ce","source":{"kind":"arxiv","id":"1704.06440","version":4},"attestation_state":"computed","paper":{"title":"Equivalence Between Policy Gradients and Soft Q-Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"John Schulman, Pieter Abbeel, Xi Chen","submitted_at":"2017-04-21T08:33:59Z","abstract_excerpt":"Two of the leading approaches for model-free reinforcement learning are policy gradient methods and $Q$-learning methods. $Q$-learning methods can be effective and sample-efficient when they work, however, it is not well-understood why they work, since empirically, the $Q$-values they estimate are very inaccurate. A partial explanation may be that $Q$-learning methods are secretly implementing policy gradient updates: we show that there is a precise equivalence between $Q$-learning and policy gradient methods in the setting of entropy-regularized reinforcement learning, that \"soft\" (entropy-re"},"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":"1704.06440","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2017-04-21T08:33:59Z","cross_cats_sorted":[],"title_canon_sha256":"d6232e35a531cd153c7eaa31b02e103c5638dbe733f4cdb0f7402562843b3390","abstract_canon_sha256":"e682ec0312b2966f5d374f8fd8944fba1fd2aa3461951d86feb5af743fc4749c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:03:25.093694Z","signature_b64":"Sf+wvu2W3AvqOBjOcVAfmEhxX3Ud2MEskcwB31t1gjM86plnw2JJsn//Jzha+/thJtp35vymF/rHAgZzRYExBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b207f83af652604df42813f763b30a3125ee0b196676c3ead9123dfe78b4f6ce","last_reissued_at":"2026-05-18T00:03:25.093186Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:03:25.093186Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Equivalence Between Policy Gradients and Soft Q-Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"John Schulman, Pieter Abbeel, Xi Chen","submitted_at":"2017-04-21T08:33:59Z","abstract_excerpt":"Two of the leading approaches for model-free reinforcement learning are policy gradient methods and $Q$-learning methods. $Q$-learning methods can be effective and sample-efficient when they work, however, it is not well-understood why they work, since empirically, the $Q$-values they estimate are very inaccurate. A partial explanation may be that $Q$-learning methods are secretly implementing policy gradient updates: we show that there is a precise equivalence between $Q$-learning and policy gradient methods in the setting of entropy-regularized reinforcement learning, that \"soft\" (entropy-re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1704.06440","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":""},"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":"1704.06440","created_at":"2026-05-18T00:03:25.093261+00:00"},{"alias_kind":"arxiv_version","alias_value":"1704.06440v4","created_at":"2026-05-18T00:03:25.093261+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1704.06440","created_at":"2026-05-18T00:03:25.093261+00:00"},{"alias_kind":"pith_short_12","alias_value":"WID7QOXWKJQE","created_at":"2026-05-18T12:31:53.515858+00:00"},{"alias_kind":"pith_short_16","alias_value":"WID7QOXWKJQE35BI","created_at":"2026-05-18T12:31:53.515858+00:00"},{"alias_kind":"pith_short_8","alias_value":"WID7QOXW","created_at":"2026-05-18T12:31:53.515858+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":14,"internal_anchor_count":6,"sample":[{"citing_arxiv_id":"2510.03508","citing_title":"D2 Actor Critic: Diffusion Actor Meets Distributional Critic","ref_index":29,"is_internal_anchor":true},{"citing_arxiv_id":"2409.02399","citing_title":"Guidance for twisted particle filter: a continuous-time perspective","ref_index":56,"is_internal_anchor":true},{"citing_arxiv_id":"2510.08141","citing_title":"SCOPE-RL: Stable and Quantitative Control of Policy Entropy in RL Post-Training","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18591","citing_title":"Randomized Advantage Transformation (RAT): Computing Natural Policy Gradients via Direct Backpropagation","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2512.15605","citing_title":"Autoregressive Language Models are Secretly Energy-Based Models: Insights into the Lookahead Capabilities of Next-Token Prediction","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2602.01505","citing_title":"Optimal Sample Complexity for Single Time-Scale Actor-Critic with Momentum","ref_index":48,"is_internal_anchor":true},{"citing_arxiv_id":"1812.05905","citing_title":"Soft Actor-Critic Algorithms and Applications","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12227","citing_title":"Combining On-Policy Optimization and Distillation for Long-Context Reasoning in Large Language Models","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"1801.01290","citing_title":"Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08007","citing_title":"Interpreting Reinforcement Learning Agents with Susceptibilities","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13780","citing_title":"Soft $Q(\\lambda)$: A multi-step off-policy method for entropy regularised reinforcement learning using eligibility traces","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14922","citing_title":"LongAct: Harnessing Intrinsic Activation Patterns for Long-Context Reinforcement Learning","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14646","citing_title":"Targeted Exploration via Unified Entropy Control for Reinforcement Learning","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19695","citing_title":"Planning in entropy-regularized Markov decision processes and games","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE","json":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE.json","graph_json":"https://pith.science/api/pith-number/WID7QOXWKJQE35BICP3WHMYKGE/graph.json","events_json":"https://pith.science/api/pith-number/WID7QOXWKJQE35BICP3WHMYKGE/events.json","paper":"https://pith.science/paper/WID7QOXW"},"agent_actions":{"view_html":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE","download_json":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE.json","view_paper":"https://pith.science/paper/WID7QOXW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1704.06440&json=true","fetch_graph":"https://pith.science/api/pith-number/WID7QOXWKJQE35BICP3WHMYKGE/graph.json","fetch_events":"https://pith.science/api/pith-number/WID7QOXWKJQE35BICP3WHMYKGE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE/action/storage_attestation","attest_author":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE/action/author_attestation","sign_citation":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE/action/citation_signature","submit_replication":"https://pith.science/pith/WID7QOXWKJQE35BICP3WHMYKGE/action/replication_record"}},"created_at":"2026-05-18T00:03:25.093261+00:00","updated_at":"2026-05-18T00:03:25.093261+00:00"}