{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5UUMGBAFWOYT34TXGLCO6NJY7Q","short_pith_number":"pith:5UUMGBAF","schema_version":"1.0","canonical_sha256":"ed28c30405b3b13df27732c4ef3538fc3d43bacc5cdce4f8951ef9a9ad127ae8","source":{"kind":"arxiv","id":"2503.05126","version":3},"attestation_state":"computed","paper":{"title":"Multi-Task Reinforcement Learning Enables Parameter Scaling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Evangelos Chatzaroulas, Isaac Woungang, Jordan Terry, Nariman Farsad, Pablo Samuel Castro, Reginald McLean","submitted_at":"2025-03-07T04:13:02Z","abstract_excerpt":"Multi-task reinforcement learning (MTRL) aims to endow a single agent with the ability to perform well on multiple tasks. Recent works have focused on developing novel sophisticated architectures to improve performance, often resulting in larger models; it is unclear, however, whether the performance gains are a consequence of the architecture design itself or the extra parameters. We argue that gains are mostly due to scale by demonstrating that naively scaling up a simple MTRL baseline to match parameter counts outperforms the more sophisticated architectures, and these gains benefit most fr"},"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":"2503.05126","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-07T04:13:02Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8a5275c6339a59a9d6e0c1c564c642341c9080cc1f8f5318130f67b09e328ca6","abstract_canon_sha256":"927fe0c2b3a8c0d8453d21d603a6c6a366bb3852ac3fbfd7605aa98ec7d4d48c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:29:36.485777Z","signature_b64":"4WxAXOG5PH8/zz0KC3H94R2RLBcdBo+aOl5YFgjpGQKcWKgaHpMmvlZrDWEDoTv6087vZerLIl7BGV5+Cl99CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ed28c30405b3b13df27732c4ef3538fc3d43bacc5cdce4f8951ef9a9ad127ae8","last_reissued_at":"2026-07-05T10:29:36.484824Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:29:36.484824Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-Task Reinforcement Learning Enables Parameter Scaling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Evangelos Chatzaroulas, Isaac Woungang, Jordan Terry, Nariman Farsad, Pablo Samuel Castro, Reginald McLean","submitted_at":"2025-03-07T04:13:02Z","abstract_excerpt":"Multi-task reinforcement learning (MTRL) aims to endow a single agent with the ability to perform well on multiple tasks. Recent works have focused on developing novel sophisticated architectures to improve performance, often resulting in larger models; it is unclear, however, whether the performance gains are a consequence of the architecture design itself or the extra parameters. We argue that gains are mostly due to scale by demonstrating that naively scaling up a simple MTRL baseline to match parameter counts outperforms the more sophisticated architectures, and these gains benefit most fr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.05126","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/2503.05126/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":"2503.05126","created_at":"2026-07-05T10:29:36.484941+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.05126v3","created_at":"2026-07-05T10:29:36.484941+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.05126","created_at":"2026-07-05T10:29:36.484941+00:00"},{"alias_kind":"pith_short_12","alias_value":"5UUMGBAFWOYT","created_at":"2026-07-05T10:29:36.484941+00:00"},{"alias_kind":"pith_short_16","alias_value":"5UUMGBAFWOYT34TX","created_at":"2026-07-05T10:29:36.484941+00:00"},{"alias_kind":"pith_short_8","alias_value":"5UUMGBAF","created_at":"2026-07-05T10:29:36.484941+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.14350","citing_title":"Distributionally Robust Multi-Task Reinforcement Learning via Adaptive Task Sampling","ref_index":210,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q","json":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q.json","graph_json":"https://pith.science/api/pith-number/5UUMGBAFWOYT34TXGLCO6NJY7Q/graph.json","events_json":"https://pith.science/api/pith-number/5UUMGBAFWOYT34TXGLCO6NJY7Q/events.json","paper":"https://pith.science/paper/5UUMGBAF"},"agent_actions":{"view_html":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q","download_json":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q.json","view_paper":"https://pith.science/paper/5UUMGBAF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.05126&json=true","fetch_graph":"https://pith.science/api/pith-number/5UUMGBAFWOYT34TXGLCO6NJY7Q/graph.json","fetch_events":"https://pith.science/api/pith-number/5UUMGBAFWOYT34TXGLCO6NJY7Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q/action/storage_attestation","attest_author":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q/action/author_attestation","sign_citation":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q/action/citation_signature","submit_replication":"https://pith.science/pith/5UUMGBAFWOYT34TXGLCO6NJY7Q/action/replication_record"}},"created_at":"2026-07-05T10:29:36.484941+00:00","updated_at":"2026-07-05T10:29:36.484941+00:00"}