{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:DMP34KVZ3TDACAIQSVAQO6ZVIH","short_pith_number":"pith:DMP34KVZ","schema_version":"1.0","canonical_sha256":"1b1fbe2ab9dcc60101109541077b3541ede70fafc4c950d812297d56fb90c5fc","source":{"kind":"arxiv","id":"2506.00592","version":1},"attestation_state":"computed","paper":{"title":"Mitigating Plasticity Loss in Continual Reinforcement Learning by Reducing Churn","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aaron Courville, Glen Berseth, Hongyao Tang, Johan Obando-Ceron, Pablo Samuel Castro","submitted_at":"2025-05-31T14:58:22Z","abstract_excerpt":"Plasticity, or the ability of an agent to adapt to new tasks, environments, or distributions, is crucial for continual learning. In this paper, we study the loss of plasticity in deep continual RL from the lens of churn: network output variability for out-of-batch data induced by mini-batch training. We demonstrate that (1) the loss of plasticity is accompanied by the exacerbation of churn due to the gradual rank decrease of the Neural Tangent Kernel (NTK) matrix; (2) reducing churn helps prevent rank collapse and adjusts the step size of regular RL gradients adaptively. Moreover, we introduce"},"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":"2506.00592","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-31T14:58:22Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"5be7f591a4555b445cc2d59367ba39cf910b2e6d72fd957346544c82ac2ab6ce","abstract_canon_sha256":"d93f503366afd03f88ff04c2226aa5e62bc3191352691c25bd06e3cd0a746f37"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:37.483816Z","signature_b64":"lewbWkn4ERJhFvLtgLeSHgDDDKGOAVG1Y3oXtLW4Z8HcaAfcyG8xNzYNOiYwNWuH/kTGTfqqqXVbw6RxfJsCDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1b1fbe2ab9dcc60101109541077b3541ede70fafc4c950d812297d56fb90c5fc","last_reissued_at":"2026-07-05T11:13:37.483334Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:37.483334Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mitigating Plasticity Loss in Continual Reinforcement Learning by Reducing Churn","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aaron Courville, Glen Berseth, Hongyao Tang, Johan Obando-Ceron, Pablo Samuel Castro","submitted_at":"2025-05-31T14:58:22Z","abstract_excerpt":"Plasticity, or the ability of an agent to adapt to new tasks, environments, or distributions, is crucial for continual learning. In this paper, we study the loss of plasticity in deep continual RL from the lens of churn: network output variability for out-of-batch data induced by mini-batch training. We demonstrate that (1) the loss of plasticity is accompanied by the exacerbation of churn due to the gradual rank decrease of the Neural Tangent Kernel (NTK) matrix; (2) reducing churn helps prevent rank collapse and adjusts the step size of regular RL gradients adaptively. Moreover, we introduce"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.00592","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/2506.00592/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":"2506.00592","created_at":"2026-07-05T11:13:37.483393+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.00592v1","created_at":"2026-07-05T11:13:37.483393+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.00592","created_at":"2026-07-05T11:13:37.483393+00:00"},{"alias_kind":"pith_short_12","alias_value":"DMP34KVZ3TDA","created_at":"2026-07-05T11:13:37.483393+00:00"},{"alias_kind":"pith_short_16","alias_value":"DMP34KVZ3TDACAIQ","created_at":"2026-07-05T11:13:37.483393+00:00"},{"alias_kind":"pith_short_8","alias_value":"DMP34KVZ","created_at":"2026-07-05T11:13:37.483393+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.12484","citing_title":"Learning, Fast and Slow: Towards LLMs That Adapt Continually","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12484","citing_title":"Learning, Fast and Slow: Towards LLMs That Adapt Continually","ref_index":59,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH","json":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH.json","graph_json":"https://pith.science/api/pith-number/DMP34KVZ3TDACAIQSVAQO6ZVIH/graph.json","events_json":"https://pith.science/api/pith-number/DMP34KVZ3TDACAIQSVAQO6ZVIH/events.json","paper":"https://pith.science/paper/DMP34KVZ"},"agent_actions":{"view_html":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH","download_json":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH.json","view_paper":"https://pith.science/paper/DMP34KVZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.00592&json=true","fetch_graph":"https://pith.science/api/pith-number/DMP34KVZ3TDACAIQSVAQO6ZVIH/graph.json","fetch_events":"https://pith.science/api/pith-number/DMP34KVZ3TDACAIQSVAQO6ZVIH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH/action/storage_attestation","attest_author":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH/action/author_attestation","sign_citation":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH/action/citation_signature","submit_replication":"https://pith.science/pith/DMP34KVZ3TDACAIQSVAQO6ZVIH/action/replication_record"}},"created_at":"2026-07-05T11:13:37.483393+00:00","updated_at":"2026-07-05T11:13:37.483393+00:00"}