{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:24TFP6W7QSWKQIVXDTX5K75QTU","short_pith_number":"pith:24TFP6W7","schema_version":"1.0","canonical_sha256":"d72657fadf84aca822b71cefd57fb09d0c39fc1600c0be45160955d787d6fcf3","source":{"kind":"arxiv","id":"2505.22642","version":3},"attestation_state":"computed","paper":{"title":"FastTD3: Simple, Fast, and Capable Reinforcement Learning for Humanoid Control","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.RO","authors_text":"Carmelo Sferrazza, Haoran Geng, Michal Nauman, Pieter Abbeel, Younggyo Seo, Zhao-Heng Yin","submitted_at":"2025-05-28T17:55:26Z","abstract_excerpt":"Reinforcement learning (RL) has driven significant progress in robotics, but its complexity and long training times remain major bottlenecks. In this report, we introduce FastTD3, a simple, fast, and capable RL algorithm that significantly speeds up training for humanoid robots in popular suites such as HumanoidBench, IsaacLab, and MuJoCo Playground. Our recipe is remarkably simple: we train an off-policy TD3 agent with several modifications -- parallel simulation, large-batch updates, a distributional critic, and carefully tuned hyperparameters. FastTD3 solves a range of HumanoidBench tasks i"},"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":"2505.22642","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2025-05-28T17:55:26Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"3233c6055baba31824e519b53c7bfc5f5684d04139fe4a5227372523f977d628","abstract_canon_sha256":"016ffde95fea532c04912d2ba4ad296f7600ded309ead2e6b971a7bff2060421"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:27.552389Z","signature_b64":"NlcdMA8fH4rFu/oIpyH5Enug/WUr58VVIYkedHyYlvvrjL9XNtIfP/ly/7rJSBqfHdxSIJF+LKJkbfwbpjt1DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d72657fadf84aca822b71cefd57fb09d0c39fc1600c0be45160955d787d6fcf3","last_reissued_at":"2026-07-05T11:13:27.551900Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:27.551900Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FastTD3: Simple, Fast, and Capable Reinforcement Learning for Humanoid Control","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.RO","authors_text":"Carmelo Sferrazza, Haoran Geng, Michal Nauman, Pieter Abbeel, Younggyo Seo, Zhao-Heng Yin","submitted_at":"2025-05-28T17:55:26Z","abstract_excerpt":"Reinforcement learning (RL) has driven significant progress in robotics, but its complexity and long training times remain major bottlenecks. In this report, we introduce FastTD3, a simple, fast, and capable RL algorithm that significantly speeds up training for humanoid robots in popular suites such as HumanoidBench, IsaacLab, and MuJoCo Playground. Our recipe is remarkably simple: we train an off-policy TD3 agent with several modifications -- parallel simulation, large-batch updates, a distributional critic, and carefully tuned hyperparameters. FastTD3 solves a range of HumanoidBench tasks i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.22642","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/2505.22642/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":"2505.22642","created_at":"2026-07-05T11:13:27.551959+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.22642v3","created_at":"2026-07-05T11:13:27.551959+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.22642","created_at":"2026-07-05T11:13:27.551959+00:00"},{"alias_kind":"pith_short_12","alias_value":"24TFP6W7QSWK","created_at":"2026-07-05T11:13:27.551959+00:00"},{"alias_kind":"pith_short_16","alias_value":"24TFP6W7QSWKQIVX","created_at":"2026-07-05T11:13:27.551959+00:00"},{"alias_kind":"pith_short_8","alias_value":"24TFP6W7","created_at":"2026-07-05T11:13:27.551959+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":18,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07830","citing_title":"Physics-Guided Biomechanical Gait Adaptation for Humanoid Locomotion on Extreme Sloped Terrains","ref_index":52,"is_internal_anchor":true},{"citing_arxiv_id":"2606.21086","citing_title":"ReFPO: Reflow Regularization for Flow Matching Policy Gradients","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05555","citing_title":"Representation Learning Enables Scalable Multitask Deep Reinforcement Learning","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31691","citing_title":"FastDSAC: Enhancing Policy Plasticity via Constrained Exploration for Scalable Humanoid Locomotion","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30072","citing_title":"ACPO: Agent-Chained Policy Optimization for Multi-Agent Reinforcement Learning","ref_index":100,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29209","citing_title":"AnyBody: Free-Form Whole-Body Humanoid Control from Arbitrary Keypoint Guidance","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30313","citing_title":"UniLab: A Heterogeneous Architecture for Robot RL Beyond GPU-Dominant Paradigms","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10236","citing_title":"When Does Non-Uniform Replay Matter in Reinforcement Learning?","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04539","citing_title":"FlashSAC: Fast and Stable Off-Policy Reinforcement Learning for High-Dimensional Robot Control","ref_index":75,"is_internal_anchor":false},{"citing_arxiv_id":"2506.15953","citing_title":"ViTacFormer: Learning Cross-Modal Representation for Visuo-Tactile Dexterous Manipulation","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2507.11019","citing_title":"Relative Entropy Pathwise Policy Optimization","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2603.12612","citing_title":"FastDSAC: Unlocking the Potential of Maximum Entropy RL in High-Dimensional Humanoid Control","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2603.15956","citing_title":"ExpertGen: Scalable Sim-to-Real Expert Policy Learning from Imperfect Behavior Priors","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10236","citing_title":"When Does Non-Uniform Replay Matter in Reinforcement Learning?","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10236","citing_title":"When Does Non-Uniform Replay Matter in Reinforcement Learning?","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25508","citing_title":"Dyna-Style Safety Augmented Reinforcement Learning: Staying Safe in the Face of Uncertainty","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06497","citing_title":"Hyperfastrl: Hypernetwork-based reinforcement learning for unified control of parametric chaotic PDEs","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04539","citing_title":"FlashSAC: Fast and Stable Off-Policy Reinforcement Learning for High-Dimensional Robot Control","ref_index":75,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU","json":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU.json","graph_json":"https://pith.science/api/pith-number/24TFP6W7QSWKQIVXDTX5K75QTU/graph.json","events_json":"https://pith.science/api/pith-number/24TFP6W7QSWKQIVXDTX5K75QTU/events.json","paper":"https://pith.science/paper/24TFP6W7"},"agent_actions":{"view_html":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU","download_json":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU.json","view_paper":"https://pith.science/paper/24TFP6W7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.22642&json=true","fetch_graph":"https://pith.science/api/pith-number/24TFP6W7QSWKQIVXDTX5K75QTU/graph.json","fetch_events":"https://pith.science/api/pith-number/24TFP6W7QSWKQIVXDTX5K75QTU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU/action/storage_attestation","attest_author":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU/action/author_attestation","sign_citation":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU/action/citation_signature","submit_replication":"https://pith.science/pith/24TFP6W7QSWKQIVXDTX5K75QTU/action/replication_record"}},"created_at":"2026-07-05T11:13:27.551959+00:00","updated_at":"2026-07-05T11:13:27.551959+00:00"}