{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2015:QOA5QNG3U2R3MSNHHC6ARRQAF2","short_pith_number":"pith:QOA5QNG3","schema_version":"1.0","canonical_sha256":"8381d834dba6a3b649a738bc08c6002ebb5fc2ddccdde70eb4bbdd2e5c82eb95","source":{"kind":"arxiv","id":"1512.08562","version":4},"attestation_state":"computed","paper":{"title":"Taming the Noise in Reinforcement Learning via Soft Updates","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IT","math.IT"],"primary_cat":"cs.LG","authors_text":"Ari Pakman, Naftali Tishby, Roy Fox","submitted_at":"2015-12-28T23:59:12Z","abstract_excerpt":"Model-free reinforcement learning algorithms, such as Q-learning, perform poorly in the early stages of learning in noisy environments, because much effort is spent unlearning biased estimates of the state-action value function. The bias results from selecting, among several noisy estimates, the apparent optimum, which may actually be suboptimal. We propose G-learning, a new off-policy learning algorithm that regularizes the value estimates by penalizing deterministic policies in the beginning of the learning process. We show that this method reduces the bias of the value-function estimation, "},"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":"1512.08562","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2015-12-28T23:59:12Z","cross_cats_sorted":["cs.IT","math.IT"],"title_canon_sha256":"c4435343ce97d059fa561682a4773276d3c57c3bfbae22dc02d77c5a4a32f5f3","abstract_canon_sha256":"fdf93c4cbea5e0303e892876d91ee27fa323c7aaf22fd5dc3e7904d7fecf7fae"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:24:45.418918Z","signature_b64":"eYm6azdp8FCIs9ulC3TUhnAZ2mPc9nMRolyCYyBZe1cSocEEhaskwD67Bg8bJcveJl9VfWXHRIcfEhMPYGx9AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8381d834dba6a3b649a738bc08c6002ebb5fc2ddccdde70eb4bbdd2e5c82eb95","last_reissued_at":"2026-05-18T00:24:45.418249Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:24:45.418249Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Taming the Noise in Reinforcement Learning via Soft Updates","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IT","math.IT"],"primary_cat":"cs.LG","authors_text":"Ari Pakman, Naftali Tishby, Roy Fox","submitted_at":"2015-12-28T23:59:12Z","abstract_excerpt":"Model-free reinforcement learning algorithms, such as Q-learning, perform poorly in the early stages of learning in noisy environments, because much effort is spent unlearning biased estimates of the state-action value function. The bias results from selecting, among several noisy estimates, the apparent optimum, which may actually be suboptimal. We propose G-learning, a new off-policy learning algorithm that regularizes the value estimates by penalizing deterministic policies in the beginning of the learning process. We show that this method reduces the bias of the value-function estimation, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1512.08562","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":"1512.08562","created_at":"2026-05-18T00:24:45.418371+00:00"},{"alias_kind":"arxiv_version","alias_value":"1512.08562v4","created_at":"2026-05-18T00:24:45.418371+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1512.08562","created_at":"2026-05-18T00:24:45.418371+00:00"},{"alias_kind":"pith_short_12","alias_value":"QOA5QNG3U2R3","created_at":"2026-05-18T12:29:37.295048+00:00"},{"alias_kind":"pith_short_16","alias_value":"QOA5QNG3U2R3MSNH","created_at":"2026-05-18T12:29:37.295048+00:00"},{"alias_kind":"pith_short_8","alias_value":"QOA5QNG3","created_at":"2026-05-18T12:29:37.295048+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2507.10797","citing_title":"Multi-Armed Sampling Problem and the End of Exploration","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2509.20265","citing_title":"Failure Modes of Maximum Entropy RLHF","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2409.12917","citing_title":"Training Language Models to Self-Correct via Reinforcement Learning","ref_index":120,"is_internal_anchor":true},{"citing_arxiv_id":"2309.00267","citing_title":"RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback","ref_index":113,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09214","citing_title":"Fast Rates for Offline Contextual Bandits with Forward-KL Regularization under Single-Policy Concentrability","ref_index":34,"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":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2","json":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2.json","graph_json":"https://pith.science/api/pith-number/QOA5QNG3U2R3MSNHHC6ARRQAF2/graph.json","events_json":"https://pith.science/api/pith-number/QOA5QNG3U2R3MSNHHC6ARRQAF2/events.json","paper":"https://pith.science/paper/QOA5QNG3"},"agent_actions":{"view_html":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2","download_json":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2.json","view_paper":"https://pith.science/paper/QOA5QNG3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1512.08562&json=true","fetch_graph":"https://pith.science/api/pith-number/QOA5QNG3U2R3MSNHHC6ARRQAF2/graph.json","fetch_events":"https://pith.science/api/pith-number/QOA5QNG3U2R3MSNHHC6ARRQAF2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2/action/storage_attestation","attest_author":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2/action/author_attestation","sign_citation":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2/action/citation_signature","submit_replication":"https://pith.science/pith/QOA5QNG3U2R3MSNHHC6ARRQAF2/action/replication_record"}},"created_at":"2026-05-18T00:24:45.418371+00:00","updated_at":"2026-05-18T00:24:45.418371+00:00"}