{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:HTKJAXRJ7DDM5J4K5AQ2NV5YP3","short_pith_number":"pith:HTKJAXRJ","schema_version":"1.0","canonical_sha256":"3cd4905e29f8c6cea78ae821a6d7b87ed71d6a09c9edfbd272b650bffb029faf","source":{"kind":"arxiv","id":"2505.10861","version":1},"attestation_state":"computed","paper":{"title":"Improving the Data-efficiency of Reinforcement Learning by Warm-starting with LLM","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chicheng Zhang, Minglai Yang, Thang Duong","submitted_at":"2025-05-16T05:03:39Z","abstract_excerpt":"We investigate the usage of Large Language Model (LLM) in collecting high-quality data to warm-start Reinforcement Learning (RL) algorithms for learning in some classical Markov Decision Process (MDP) environments. In this work, we focus on using LLM to generate an off-policy dataset that sufficiently covers state-actions visited by optimal policies, then later using an RL algorithm to explore the environment and improve the policy suggested by the LLM. Our algorithm, LORO, can both converge to an optimal policy and have a high sample efficiency thanks to the LLM's good starting policy. On mul"},"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":"2505.10861","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-16T05:03:39Z","cross_cats_sorted":[],"title_canon_sha256":"b125237f2c081fbbe14a92b9723c0dd6ec951110ea9bf3f2bec22eb32a50ca62","abstract_canon_sha256":"d4415b995f495480b5bcd583b18bb9baa34e36a89db0d438fb6020f646906681"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:04:07.231269Z","signature_b64":"FcsuDvZCG9tJ3j7ykJnPVc3jjhyPKAogm40qB7Tyygpmfp9BK+ALl3XsO7gTo4Mdk3JWPGr6JFLFvDKByLrfBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3cd4905e29f8c6cea78ae821a6d7b87ed71d6a09c9edfbd272b650bffb029faf","last_reissued_at":"2026-07-05T11:04:07.230782Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:04:07.230782Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving the Data-efficiency of Reinforcement Learning by Warm-starting with LLM","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chicheng Zhang, Minglai Yang, Thang Duong","submitted_at":"2025-05-16T05:03:39Z","abstract_excerpt":"We investigate the usage of Large Language Model (LLM) in collecting high-quality data to warm-start Reinforcement Learning (RL) algorithms for learning in some classical Markov Decision Process (MDP) environments. In this work, we focus on using LLM to generate an off-policy dataset that sufficiently covers state-actions visited by optimal policies, then later using an RL algorithm to explore the environment and improve the policy suggested by the LLM. Our algorithm, LORO, can both converge to an optimal policy and have a high sample efficiency thanks to the LLM's good starting policy. On mul"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.10861","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/2505.10861/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":"2505.10861","created_at":"2026-07-05T11:04:07.230842+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.10861v1","created_at":"2026-07-05T11:04:07.230842+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.10861","created_at":"2026-07-05T11:04:07.230842+00:00"},{"alias_kind":"pith_short_12","alias_value":"HTKJAXRJ7DDM","created_at":"2026-07-05T11:04:07.230842+00:00"},{"alias_kind":"pith_short_16","alias_value":"HTKJAXRJ7DDM5J4K","created_at":"2026-07-05T11:04:07.230842+00:00"},{"alias_kind":"pith_short_8","alias_value":"HTKJAXRJ","created_at":"2026-07-05T11:04:07.230842+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.06015","citing_title":"ProDVI: Programmatic Dynamics Priors for Value Network Initialization","ref_index":62,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3","json":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3.json","graph_json":"https://pith.science/api/pith-number/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/graph.json","events_json":"https://pith.science/api/pith-number/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/events.json","paper":"https://pith.science/paper/HTKJAXRJ"},"agent_actions":{"view_html":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3","download_json":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3.json","view_paper":"https://pith.science/paper/HTKJAXRJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.10861&json=true","fetch_graph":"https://pith.science/api/pith-number/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/graph.json","fetch_events":"https://pith.science/api/pith-number/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/action/storage_attestation","attest_author":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/action/author_attestation","sign_citation":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/action/citation_signature","submit_replication":"https://pith.science/pith/HTKJAXRJ7DDM5J4K5AQ2NV5YP3/action/replication_record"}},"created_at":"2026-07-05T11:04:07.230842+00:00","updated_at":"2026-07-05T11:04:07.230842+00:00"}