{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KJW4776S5RZHZVRO2JPYGA2UV3","short_pith_number":"pith:KJW4776S","schema_version":"1.0","canonical_sha256":"526dcfffd2ec727cd62ed25f830354aee01f06e0177613bbf5c675d5181860b1","source":{"kind":"arxiv","id":"2602.01460","version":3},"attestation_state":"computed","paper":{"title":"Non-Uniform Noise-to-Signal Ratio in the REINFORCE Policy-Gradient Estimator","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"math.OC","authors_text":"Haoyu Han, Heng Yang","submitted_at":"2026-02-01T22:05:48Z","abstract_excerpt":"Policy-gradient methods are widely used in reinforcement learning, yet training often becomes unstable or slows down as learning progresses. We study this phenomenon through the noise-to-signal ratio (NSR) of a policy-gradient estimator, defined as the estimator variance (noise) normalized by the squared norm of the true gradient (signal). Our main result is that, for (i) finite-horizon linear systems with Gaussian policies and linear state-feedback, and (ii) finite-horizon polynomial systems with Gaussian policies and polynomial feedback, the NSR of the REINFORCE estimator can be characterize"},"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":"2602.01460","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"math.OC","submitted_at":"2026-02-01T22:05:48Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"44ad21add2c04ce090f744193bb691c34220ae700db90d15e3fc500a23f2a3f5","abstract_canon_sha256":"686353101109378fbc7c47cb76324391b3d9fa11a997886ff90c2b5114481b0f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:03:15.568780Z","signature_b64":"nnyo20JZ1OKnkVdNaQ6UrxWHbyPhICm+vm60jcbyMRqZ+G5fhKQFpnijXMwhiuYIlYBouxH3U5CltdrOUdFICA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"526dcfffd2ec727cd62ed25f830354aee01f06e0177613bbf5c675d5181860b1","last_reissued_at":"2026-06-02T01:03:15.568227Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:03:15.568227Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Non-Uniform Noise-to-Signal Ratio in the REINFORCE Policy-Gradient Estimator","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"math.OC","authors_text":"Haoyu Han, Heng Yang","submitted_at":"2026-02-01T22:05:48Z","abstract_excerpt":"Policy-gradient methods are widely used in reinforcement learning, yet training often becomes unstable or slows down as learning progresses. We study this phenomenon through the noise-to-signal ratio (NSR) of a policy-gradient estimator, defined as the estimator variance (noise) normalized by the squared norm of the true gradient (signal). Our main result is that, for (i) finite-horizon linear systems with Gaussian policies and linear state-feedback, and (ii) finite-horizon polynomial systems with Gaussian policies and polynomial feedback, the NSR of the REINFORCE estimator can be characterize"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.01460","kind":"arxiv","version":3},"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/2602.01460/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":"2602.01460","created_at":"2026-06-02T01:03:15.568284+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.01460v3","created_at":"2026-06-02T01:03:15.568284+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.01460","created_at":"2026-06-02T01:03:15.568284+00:00"},{"alias_kind":"pith_short_12","alias_value":"KJW4776S5RZH","created_at":"2026-06-02T01:03:15.568284+00:00"},{"alias_kind":"pith_short_16","alias_value":"KJW4776S5RZHZVRO","created_at":"2026-06-02T01:03:15.568284+00:00"},{"alias_kind":"pith_short_8","alias_value":"KJW4776S","created_at":"2026-06-02T01:03:15.568284+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2604.21456","citing_title":"Tempered Sequential Monte Carlo for Trajectory and Policy Optimization with Differentiable Dynamics","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06732","citing_title":"On Training in Imagination","ref_index":36,"is_internal_anchor":true},{"citing_arxiv_id":"2604.21456","citing_title":"Tempered Sequential Monte Carlo for Trajectory and Policy Optimization with Differentiable Dynamics","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2605.01865","citing_title":"Quality-Aware Exploration Budget Allocation for Cooperative Multi-Agent Reinforcement Learning","ref_index":45,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3","json":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3.json","graph_json":"https://pith.science/api/pith-number/KJW4776S5RZHZVRO2JPYGA2UV3/graph.json","events_json":"https://pith.science/api/pith-number/KJW4776S5RZHZVRO2JPYGA2UV3/events.json","paper":"https://pith.science/paper/KJW4776S"},"agent_actions":{"view_html":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3","download_json":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3.json","view_paper":"https://pith.science/paper/KJW4776S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.01460&json=true","fetch_graph":"https://pith.science/api/pith-number/KJW4776S5RZHZVRO2JPYGA2UV3/graph.json","fetch_events":"https://pith.science/api/pith-number/KJW4776S5RZHZVRO2JPYGA2UV3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3/action/storage_attestation","attest_author":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3/action/author_attestation","sign_citation":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3/action/citation_signature","submit_replication":"https://pith.science/pith/KJW4776S5RZHZVRO2JPYGA2UV3/action/replication_record"}},"created_at":"2026-06-02T01:03:15.568284+00:00","updated_at":"2026-06-02T01:03:15.568284+00:00"}