{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:XQDMVRWSUSCFSL4TG7OUUAG7UP","short_pith_number":"pith:XQDMVRWS","schema_version":"1.0","canonical_sha256":"bc06cac6d2a484592f9337dd4a00dfa3d941bcf6afe1879bc0195de99a3ebc1b","source":{"kind":"arxiv","id":"2003.13661","version":2},"attestation_state":"computed","paper":{"title":"Multi-Task Reinforcement Learning with Soft Modularization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.RO","stat.ML"],"primary_cat":"cs.LG","authors_text":"Huazhe Xu, Ruihan Yang, Xiaolong Wang, Yi Wu","submitted_at":"2020-03-30T17:47:04Z","abstract_excerpt":"Multi-task learning is a very challenging problem in reinforcement learning. While training multiple tasks jointly allow the policies to share parameters across different tasks, the optimization problem becomes non-trivial: It remains unclear what parameters in the network should be reused across tasks, and how the gradients from different tasks may interfere with each other. Thus, instead of naively sharing parameters across tasks, we introduce an explicit modularization technique on policy representation to alleviate this optimization issue. Given a base policy network, we design a routing n"},"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":"2003.13661","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-03-30T17:47:04Z","cross_cats_sorted":["cs.AI","cs.RO","stat.ML"],"title_canon_sha256":"759565234e253fb42757e9a96af08223e2e92d7be8a396c0b9847d50741fb3c6","abstract_canon_sha256":"0eb401bc26bc5e37457c9cd794d1c5245d2809db255f8bbbb1ffb61718b620a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:57:15.668815Z","signature_b64":"AI/aw/RIoVVtfYM4UBGvOC/HrAd+DR07YhjrV34ET415ivWqnuzm5tMp1bvI80PamciD3Hs+PM/vJa6EFI4XBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc06cac6d2a484592f9337dd4a00dfa3d941bcf6afe1879bc0195de99a3ebc1b","last_reissued_at":"2026-07-05T01:57:15.668215Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:57:15.668215Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-Task Reinforcement Learning with Soft Modularization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.RO","stat.ML"],"primary_cat":"cs.LG","authors_text":"Huazhe Xu, Ruihan Yang, Xiaolong Wang, Yi Wu","submitted_at":"2020-03-30T17:47:04Z","abstract_excerpt":"Multi-task learning is a very challenging problem in reinforcement learning. While training multiple tasks jointly allow the policies to share parameters across different tasks, the optimization problem becomes non-trivial: It remains unclear what parameters in the network should be reused across tasks, and how the gradients from different tasks may interfere with each other. Thus, instead of naively sharing parameters across tasks, we introduce an explicit modularization technique on policy representation to alleviate this optimization issue. Given a base policy network, we design a routing n"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2003.13661","kind":"arxiv","version":2},"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/2003.13661/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":"2003.13661","created_at":"2026-07-05T01:57:15.668288+00:00"},{"alias_kind":"arxiv_version","alias_value":"2003.13661v2","created_at":"2026-07-05T01:57:15.668288+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2003.13661","created_at":"2026-07-05T01:57:15.668288+00:00"},{"alias_kind":"pith_short_12","alias_value":"XQDMVRWSUSCF","created_at":"2026-07-05T01:57:15.668288+00:00"},{"alias_kind":"pith_short_16","alias_value":"XQDMVRWSUSCFSL4T","created_at":"2026-07-05T01:57:15.668288+00:00"},{"alias_kind":"pith_short_8","alias_value":"XQDMVRWS","created_at":"2026-07-05T01:57:15.668288+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.11473","citing_title":"TOPPO: Rethinking PPO for Multi-Task Reinforcement Learning with Critic Balancing","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP","json":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP.json","graph_json":"https://pith.science/api/pith-number/XQDMVRWSUSCFSL4TG7OUUAG7UP/graph.json","events_json":"https://pith.science/api/pith-number/XQDMVRWSUSCFSL4TG7OUUAG7UP/events.json","paper":"https://pith.science/paper/XQDMVRWS"},"agent_actions":{"view_html":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP","download_json":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP.json","view_paper":"https://pith.science/paper/XQDMVRWS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2003.13661&json=true","fetch_graph":"https://pith.science/api/pith-number/XQDMVRWSUSCFSL4TG7OUUAG7UP/graph.json","fetch_events":"https://pith.science/api/pith-number/XQDMVRWSUSCFSL4TG7OUUAG7UP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP/action/storage_attestation","attest_author":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP/action/author_attestation","sign_citation":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP/action/citation_signature","submit_replication":"https://pith.science/pith/XQDMVRWSUSCFSL4TG7OUUAG7UP/action/replication_record"}},"created_at":"2026-07-05T01:57:15.668288+00:00","updated_at":"2026-07-05T01:57:15.668288+00:00"}