{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:YFE5TSVZH4IPEPUZTYSPMOWQBB","short_pith_number":"pith:YFE5TSVZ","schema_version":"1.0","canonical_sha256":"c149d9cab93f10f23e999e24f63ad0085b35c2e67bb291a0a272b4607f96acc1","source":{"kind":"arxiv","id":"2402.08609","version":3},"attestation_state":"computed","paper":{"title":"Mixtures of Experts Unlock Parameter Scaling for Deep RL","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Clare Lyle, Doina Precup, Ghada Sokar, Gintare Karolina Dziugaite, Jakob Foerster, Jesse Farebrother, Johan Obando-Ceron, Pablo Samuel Castro, Timon Willi","submitted_at":"2024-02-13T17:18:56Z","abstract_excerpt":"The recent rapid progress in (self) supervised learning models is in large part predicted by empirical scaling laws: a model's performance scales proportionally to its size. Analogous scaling laws remain elusive for reinforcement learning domains, however, where increasing the parameter count of a model often hurts its final performance. In this paper, we demonstrate that incorporating Mixture-of-Expert (MoE) modules, and in particular Soft MoEs (Puigcerver et al., 2023), into value-based networks results in more parameter-scalable models, evidenced by substantial performance increases across "},"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":"2402.08609","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-13T17:18:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"5fd953777ae88d99cc1dc41e5fa5be2a8fd8f0b893627c6accb5efe7107cdc59","abstract_canon_sha256":"33e0a16c29754bbff23e1b7707730ae8c869e519e6f226ff49e8c4a3b800d402"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:37:04.803527Z","signature_b64":"eNB2KmEKpORilgXFkDrBFliYIprS454d19Fopd7GMKy2vsKuZhk5bYqGr3srHelghWAyYAoXcvIXbmkTJyDUCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c149d9cab93f10f23e999e24f63ad0085b35c2e67bb291a0a272b4607f96acc1","last_reissued_at":"2026-07-05T08:37:04.803009Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:37:04.803009Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mixtures of Experts Unlock Parameter Scaling for Deep RL","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Clare Lyle, Doina Precup, Ghada Sokar, Gintare Karolina Dziugaite, Jakob Foerster, Jesse Farebrother, Johan Obando-Ceron, Pablo Samuel Castro, Timon Willi","submitted_at":"2024-02-13T17:18:56Z","abstract_excerpt":"The recent rapid progress in (self) supervised learning models is in large part predicted by empirical scaling laws: a model's performance scales proportionally to its size. Analogous scaling laws remain elusive for reinforcement learning domains, however, where increasing the parameter count of a model often hurts its final performance. In this paper, we demonstrate that incorporating Mixture-of-Expert (MoE) modules, and in particular Soft MoEs (Puigcerver et al., 2023), into value-based networks results in more parameter-scalable models, evidenced by substantial performance increases across "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.08609","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/2402.08609/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":"2402.08609","created_at":"2026-07-05T08:37:04.803079+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.08609v3","created_at":"2026-07-05T08:37:04.803079+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.08609","created_at":"2026-07-05T08:37:04.803079+00:00"},{"alias_kind":"pith_short_12","alias_value":"YFE5TSVZH4IP","created_at":"2026-07-05T08:37:04.803079+00:00"},{"alias_kind":"pith_short_16","alias_value":"YFE5TSVZH4IPEPUZ","created_at":"2026-07-05T08:37:04.803079+00:00"},{"alias_kind":"pith_short_8","alias_value":"YFE5TSVZ","created_at":"2026-07-05T08:37:04.803079+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20411","citing_title":"Direct Advantage Estimation for Scalable and Sample-efficient Deep Reinforcement Learning","ref_index":88,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21557","citing_title":"Scalable Reinforcement Learning via Adaptive Batch Scaling","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27722","citing_title":"NUCLEUS-MoE: Unified Model of Pool Boiling for Liquid Cooling","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21557","citing_title":"Scalable Reinforcement Learning via Adaptive Batch Scaling","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2410.12557","citing_title":"One Step Diffusion via Shortcut Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12909","citing_title":"Tree Learning: A Multi-Skill Continual Learning Framework for Humanoid Robots","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB","json":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB.json","graph_json":"https://pith.science/api/pith-number/YFE5TSVZH4IPEPUZTYSPMOWQBB/graph.json","events_json":"https://pith.science/api/pith-number/YFE5TSVZH4IPEPUZTYSPMOWQBB/events.json","paper":"https://pith.science/paper/YFE5TSVZ"},"agent_actions":{"view_html":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB","download_json":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB.json","view_paper":"https://pith.science/paper/YFE5TSVZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.08609&json=true","fetch_graph":"https://pith.science/api/pith-number/YFE5TSVZH4IPEPUZTYSPMOWQBB/graph.json","fetch_events":"https://pith.science/api/pith-number/YFE5TSVZH4IPEPUZTYSPMOWQBB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB/action/storage_attestation","attest_author":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB/action/author_attestation","sign_citation":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB/action/citation_signature","submit_replication":"https://pith.science/pith/YFE5TSVZH4IPEPUZTYSPMOWQBB/action/replication_record"}},"created_at":"2026-07-05T08:37:04.803079+00:00","updated_at":"2026-07-05T08:37:04.803079+00:00"}