{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2016:JBYBPTHUX4OUYCS63SMXM4KSPG","short_pith_number":"pith:JBYBPTHU","schema_version":"1.0","canonical_sha256":"487017ccf4bf1d4c0a5edc997671527980eee3496639585bdb2d03df717a25d0","source":{"kind":"arxiv","id":"1611.09321","version":3},"attestation_state":"computed","paper":{"title":"Improving Policy Gradient by Exploring Under-appreciated Rewards","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Dale Schuurmans, Mohammad Norouzi, Ofir Nachum","submitted_at":"2016-11-28T20:15:55Z","abstract_excerpt":"This paper presents a novel form of policy gradient for model-free reinforcement learning (RL) with improved exploration properties. Current policy-based methods use entropy regularization to encourage undirected exploration of the reward landscape, which is ineffective in high dimensional spaces with sparse rewards. We propose a more directed exploration strategy that promotes exploration of under-appreciated reward regions. An action sequence is considered under-appreciated if its log-probability under the current policy under-estimates its resulting reward. The proposed exploration strategy"},"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":"1611.09321","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2016-11-28T20:15:55Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"28e34a4949930c4f2ef49dc599c76b635d7a20b7ed3b89042b46a754659a065c","abstract_canon_sha256":"4450693b8d02b7735347d9d67986d95570b8340c71a72da37b5d5d51cc9b255d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:48:36.943051Z","signature_b64":"/Fr0qNSZdhRjO33MXAgNM+Qn5G/dmnqSXEEJCVM0YwGMMxlDLWwnDCMo4GRiTjKuWAg9paDM4YSaxsOkqeicBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"487017ccf4bf1d4c0a5edc997671527980eee3496639585bdb2d03df717a25d0","last_reissued_at":"2026-05-18T00:48:36.942455Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:48:36.942455Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving Policy Gradient by Exploring Under-appreciated Rewards","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Dale Schuurmans, Mohammad Norouzi, Ofir Nachum","submitted_at":"2016-11-28T20:15:55Z","abstract_excerpt":"This paper presents a novel form of policy gradient for model-free reinforcement learning (RL) with improved exploration properties. Current policy-based methods use entropy regularization to encourage undirected exploration of the reward landscape, which is ineffective in high dimensional spaces with sparse rewards. We propose a more directed exploration strategy that promotes exploration of under-appreciated reward regions. An action sequence is considered under-appreciated if its log-probability under the current policy under-estimates its resulting reward. The proposed exploration strategy"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1611.09321","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":""},"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":"1611.09321","created_at":"2026-05-18T00:48:36.942544+00:00"},{"alias_kind":"arxiv_version","alias_value":"1611.09321v3","created_at":"2026-05-18T00:48:36.942544+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1611.09321","created_at":"2026-05-18T00:48:36.942544+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2409.12917","citing_title":"Training Language Models to Self-Correct via Reinforcement Learning","ref_index":224,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG","json":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG.json","graph_json":"https://pith.science/api/pith-number/JBYBPTHUX4OUYCS63SMXM4KSPG/graph.json","events_json":"https://pith.science/api/pith-number/JBYBPTHUX4OUYCS63SMXM4KSPG/events.json","paper":"https://pith.science/paper/JBYBPTHU"},"agent_actions":{"view_html":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG","download_json":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG.json","view_paper":"https://pith.science/paper/JBYBPTHU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1611.09321&json=true","fetch_graph":"https://pith.science/api/pith-number/JBYBPTHUX4OUYCS63SMXM4KSPG/graph.json","fetch_events":"https://pith.science/api/pith-number/JBYBPTHUX4OUYCS63SMXM4KSPG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG/action/storage_attestation","attest_author":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG/action/author_attestation","sign_citation":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG/action/citation_signature","submit_replication":"https://pith.science/pith/JBYBPTHUX4OUYCS63SMXM4KSPG/action/replication_record"}},"created_at":"2026-05-18T00:48:36.942544+00:00","updated_at":"2026-05-18T00:48:36.942544+00:00"}