{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:4RANQRIG7CIP6ZY4O2KYPT54NN","short_pith_number":"pith:4RANQRIG","schema_version":"1.0","canonical_sha256":"e440d84506f890ff671c769587cfbc6b63cefa2aa6492989b34337cb0c3515af","source":{"kind":"arxiv","id":"2310.12931","version":2},"attestation_state":"computed","paper":{"title":"Eureka: Human-Level Reward Design via Coding Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Large language models can design reward functions for robot tasks that outperform those created by human experts.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.RO","authors_text":"Anima Anandkumar, De-An Huang, Dinesh Jayaraman, Guanzhi Wang, Linxi Fan, Osbert Bastani, William Liang, Yecheng Jason Ma, Yuke Zhu","submitted_at":"2023-10-19T17:31:01Z","abstract_excerpt":"Large Language Models (LLMs) have excelled as high-level semantic planners for sequential decision-making tasks. However, harnessing them to learn complex low-level manipulation tasks, such as dexterous pen spinning, remains an open problem. We bridge this fundamental gap and present Eureka, a human-level reward design algorithm powered by LLMs. Eureka exploits the remarkable zero-shot generation, code-writing, and in-context improvement capabilities of state-of-the-art LLMs, such as GPT-4, to perform evolutionary optimization over reward code. The resulting rewards can then be used to acquire"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2310.12931","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2023-10-19T17:31:01Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"cb5b85979677dc841ed11123093ae61df2a02fa38dcba4fe26ffe246c855c92e","abstract_canon_sha256":"ad3784548ded4a14b5fafc6dbc128558fb41c2cdec7eae24949d70c5961dbb6c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:21.856546Z","signature_b64":"1Xq3VKjvDfXAs4cf0HHgQvEoKAAmg6zno7I5qsNurEweK39ebaTpct3exCEcM/LP2ICtNAb9FbCSJie3dUImDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e440d84506f890ff671c769587cfbc6b63cefa2aa6492989b34337cb0c3515af","last_reissued_at":"2026-05-17T23:39:21.855779Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:21.855779Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Eureka: Human-Level Reward Design via Coding Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Large language models can design reward functions for robot tasks that outperform those created by human experts.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.RO","authors_text":"Anima Anandkumar, De-An Huang, Dinesh Jayaraman, Guanzhi Wang, Linxi Fan, Osbert Bastani, William Liang, Yecheng Jason Ma, Yuke Zhu","submitted_at":"2023-10-19T17:31:01Z","abstract_excerpt":"Large Language Models (LLMs) have excelled as high-level semantic planners for sequential decision-making tasks. However, harnessing them to learn complex low-level manipulation tasks, such as dexterous pen spinning, remains an open problem. We bridge this fundamental gap and present Eureka, a human-level reward design algorithm powered by LLMs. Eureka exploits the remarkable zero-shot generation, code-writing, and in-context improvement capabilities of state-of-the-art LLMs, such as GPT-4, to perform evolutionary optimization over reward code. The resulting rewards can then be used to acquire"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Without any task-specific prompting or pre-defined reward templates, Eureka generates reward functions that outperform expert human-engineered rewards. In a diverse suite of 29 open-source RL environments that include 10 distinct robot morphologies, Eureka outperforms human experts on 83% of the tasks, leading to an average normalized improvement of 52%.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That LLM-generated reward code will continue to produce stable, non-degenerate policies when transferred to new tasks or real hardware, rather than exploiting simulator quirks that do not appear in the reported 29 environments.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Eureka uses LLMs for evolutionary optimization of reward code to outperform human experts on 83% of 29 RL tasks with 52% average improvement and enables gradient-free RLHF.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Large language models can design reward functions for robot tasks that outperform those created by human experts.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"400320205c025bb9e78d1d801867ebf301ca5aefe735e9adec57308f006a3542"},"source":{"id":"2310.12931","kind":"arxiv","version":2},"verdict":{"id":"521bdc28-8cd4-4e95-be37-5921a4011a8c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T20:10:45.963363Z","strongest_claim":"Without any task-specific prompting or pre-defined reward templates, Eureka generates reward functions that outperform expert human-engineered rewards. In a diverse suite of 29 open-source RL environments that include 10 distinct robot morphologies, Eureka outperforms human experts on 83% of the tasks, leading to an average normalized improvement of 52%.","one_line_summary":"Eureka uses LLMs for evolutionary optimization of reward code to outperform human experts on 83% of 29 RL tasks with 52% average improvement and enables gradient-free RLHF.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That LLM-generated reward code will continue to produce stable, non-degenerate policies when transferred to new tasks or real hardware, rather than exploiting simulator quirks that do not appear in the reported 29 environments.","pith_extraction_headline":"Large language models can design reward functions for robot tasks that outperform those created by human experts."},"references":{"count":14,"sample":[{"doi":"","year":null,"title":"If you see phrases like [NUM: default_value], replace the entire phrase with a numerical value","work_id":"e1404de5-7f99-4b50-b730-b8dd755a9245","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"If you see phrases like {CHOICE: choice1, choice2, ...}, it means you should replace the entire phrase with one of the choices listed","work_id":"08d69275-c02c-44ab-ac69-f6ba9199ff18","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"If you see [optional], it means you only add that line if necessary for the task, otherwise remove that line","work_id":"cd6a8140-9399-4f50-b78d-ffda67ed8f68","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Do not invent new objects not listed here","work_id":"9aa3520b-1406-4dc3-b23d-d4453816fbc5","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Always start the description with [start of plan] and end it with [end of plan]","work_id":"76f847c9-fc23-4dda-927a-262b3d2c5909","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":14,"snapshot_sha256":"e0f3eb979b31a037bc6b19dae15be2ca86d18273a46c2bba621f617a7a72af06","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f69cc1d530b44f7709403e4ba609fffa2a572d54bca81a04ba0d6f8219c438f8"},"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":"2310.12931","created_at":"2026-05-17T23:39:21.855905+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.12931v2","created_at":"2026-05-17T23:39:21.855905+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.12931","created_at":"2026-05-17T23:39:21.855905+00:00"},{"alias_kind":"pith_short_12","alias_value":"4RANQRIG7CIP","created_at":"2026-05-18T12:33:33.725879+00:00"},{"alias_kind":"pith_short_16","alias_value":"4RANQRIG7CIP6ZY4","created_at":"2026-05-18T12:33:33.725879+00:00"},{"alias_kind":"pith_short_8","alias_value":"4RANQRIG","created_at":"2026-05-18T12:33:33.725879+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":37,"internal_anchor_count":37,"sample":[{"citing_arxiv_id":"2605.22123","citing_title":"Beyond Pixels: Learning Invariant Rewards for Real-World Robotics From a Few Demonstrations","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2603.12145","citing_title":"Automatic Generation of High-Performance RL Environments","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07358","citing_title":"A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applications","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2506.04565","citing_title":"From Standalone LLMs to Integrated Intelligence: A Survey of Compound Al Systems","ref_index":118,"is_internal_anchor":true},{"citing_arxiv_id":"2509.00338","citing_title":"Scalable Option Learning in High-Throughput Environments","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2509.16615","citing_title":"LLM-Guided Task- and Affordance-Level Exploration in Reinforcement Learning","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2401.03568","citing_title":"Agent AI: Surveying the Horizons of Multimodal Interaction","ref_index":149,"is_internal_anchor":true},{"citing_arxiv_id":"2411.04983","citing_title":"DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2509.19349","citing_title":"ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolution","ref_index":238,"is_internal_anchor":true},{"citing_arxiv_id":"2503.22020","citing_title":"CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2602.18142","citing_title":"Toward Automated Virtual Electronic Control Unit (ECU) Twins for Shift-Left Automotive Software Testing","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2509.09674","citing_title":"SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2602.20867","citing_title":"SoK: Agentic Skills -- Beyond Tool Use in LLM Agents","ref_index":47,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09423","citing_title":"SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning","ref_index":54,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26969","citing_title":"AgenticRecTune: Multi-Agent with Self-Evolving Skillhub for Recommendation System Optimization","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02911","citing_title":"Learning Task-Invariant Properties via Dreamer: Enabling Efficient Policy Transfer for Quadruped Robots","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2307.05973","citing_title":"VoxPoser: Composable 3D Value Maps for Robotic Manipulation with Language Models","ref_index":84,"is_internal_anchor":true},{"citing_arxiv_id":"2605.02600","citing_title":"CoRAL: Contact-Rich Adaptive LLM-based Control for Robotic Manipulation","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11859","citing_title":"EvoNav: Evolutionary Reward Function Design for Robot Navigation with Large Language Models","ref_index":29,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12421","citing_title":"Formalize, Don't Optimize: The Heuristic Trap in LLM-Generated Combinatorial Solvers","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11359","citing_title":"CVEvolve: Autonomous Algorithm Discovery for Unstructured Scientific Data Processing","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10118","citing_title":"Plan in Sandbox, Navigate in Open Worlds: Learning Physics-Grounded Abstracted Experience for Embodied Navigation","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09423","citing_title":"SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning","ref_index":54,"is_internal_anchor":true},{"citing_arxiv_id":"2604.24043","citing_title":"A2DEPT: Large Language Model-Driven Automated Algorithm Design via Evolutionary Program Trees","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26969","citing_title":"AgenticRecTune: Multi-Agent with Self-Evolving Skillhub for Recommendation System Optimization","ref_index":15,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN","json":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN.json","graph_json":"https://pith.science/api/pith-number/4RANQRIG7CIP6ZY4O2KYPT54NN/graph.json","events_json":"https://pith.science/api/pith-number/4RANQRIG7CIP6ZY4O2KYPT54NN/events.json","paper":"https://pith.science/paper/4RANQRIG"},"agent_actions":{"view_html":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN","download_json":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN.json","view_paper":"https://pith.science/paper/4RANQRIG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.12931&json=true","fetch_graph":"https://pith.science/api/pith-number/4RANQRIG7CIP6ZY4O2KYPT54NN/graph.json","fetch_events":"https://pith.science/api/pith-number/4RANQRIG7CIP6ZY4O2KYPT54NN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN/action/storage_attestation","attest_author":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN/action/author_attestation","sign_citation":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN/action/citation_signature","submit_replication":"https://pith.science/pith/4RANQRIG7CIP6ZY4O2KYPT54NN/action/replication_record"}},"created_at":"2026-05-17T23:39:21.855905+00:00","updated_at":"2026-05-17T23:39:21.855905+00:00"}