{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:Z4CGJHK5XDYLLS7QSLH4IGGQ7F","short_pith_number":"pith:Z4CGJHK5","schema_version":"1.0","canonical_sha256":"cf04649d5db8f0b5cbf092cfc418d0f954d3590b43a7c411cd83406e5b3665f9","source":{"kind":"arxiv","id":"2503.03438","version":1},"attestation_state":"computed","paper":{"title":"Gradient Deconfliction via Orthogonal Projections onto Subspaces For Multi-task Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bo Zheng, Hongbo Deng, Hui Zhao, Jian Xu, Pengjie Wang, Shijie Zhu, Tianshu Wu","submitted_at":"2025-03-05T12:13:08Z","abstract_excerpt":"Although multi-task learning (MTL) has been a preferred approach and successfully applied in many real-world scenarios, MTL models are not guaranteed to outperform single-task models on all tasks mainly due to the negative effects of conflicting gradients among the tasks. In this paper, we fully examine the influence of conflicting gradients and further emphasize the importance and advantages of achieving non-conflicting gradients which allows simple but effective trade-off strategies among the tasks with stable performance. Based on our findings, we propose the Gradient Deconfliction via Orth"},"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":"2503.03438","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-05T12:13:08Z","cross_cats_sorted":[],"title_canon_sha256":"245061ad79e5ac579a5fd5172fb7cd0ee9c0f2f4f3523cb43d6c20c9ceacd25a","abstract_canon_sha256":"65b1b7f46c11449ea5df370dd2f9c20da5030c72aa4ea0bdf193816860798226"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:24:45.469244Z","signature_b64":"bzJisjX6TT3Eaiq42yAjq9ElcnxnyeZVp4UO8yOz/hUifi0AjJX0Mb17y1VFPaGl6id5GXa/vu7ldmF905XBBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cf04649d5db8f0b5cbf092cfc418d0f954d3590b43a7c411cd83406e5b3665f9","last_reissued_at":"2026-07-05T10:24:45.468820Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:24:45.468820Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Gradient Deconfliction via Orthogonal Projections onto Subspaces For Multi-task Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bo Zheng, Hongbo Deng, Hui Zhao, Jian Xu, Pengjie Wang, Shijie Zhu, Tianshu Wu","submitted_at":"2025-03-05T12:13:08Z","abstract_excerpt":"Although multi-task learning (MTL) has been a preferred approach and successfully applied in many real-world scenarios, MTL models are not guaranteed to outperform single-task models on all tasks mainly due to the negative effects of conflicting gradients among the tasks. In this paper, we fully examine the influence of conflicting gradients and further emphasize the importance and advantages of achieving non-conflicting gradients which allows simple but effective trade-off strategies among the tasks with stable performance. Based on our findings, we propose the Gradient Deconfliction via Orth"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.03438","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/2503.03438/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":"2503.03438","created_at":"2026-07-05T10:24:45.468883+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.03438v1","created_at":"2026-07-05T10:24:45.468883+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.03438","created_at":"2026-07-05T10:24:45.468883+00:00"},{"alias_kind":"pith_short_12","alias_value":"Z4CGJHK5XDYL","created_at":"2026-07-05T10:24:45.468883+00:00"},{"alias_kind":"pith_short_16","alias_value":"Z4CGJHK5XDYLLS7Q","created_at":"2026-07-05T10:24:45.468883+00:00"},{"alias_kind":"pith_short_8","alias_value":"Z4CGJHK5","created_at":"2026-07-05T10:24:45.468883+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18688","citing_title":"Dual-Channel Grounded World Modeling (DCGWM): Structural Prevention of Objective Interference Collapse via Heterogeneous External Grounding with Inward-Only Gradient Flow","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F","json":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F.json","graph_json":"https://pith.science/api/pith-number/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/graph.json","events_json":"https://pith.science/api/pith-number/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/events.json","paper":"https://pith.science/paper/Z4CGJHK5"},"agent_actions":{"view_html":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F","download_json":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F.json","view_paper":"https://pith.science/paper/Z4CGJHK5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.03438&json=true","fetch_graph":"https://pith.science/api/pith-number/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/graph.json","fetch_events":"https://pith.science/api/pith-number/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/action/storage_attestation","attest_author":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/action/author_attestation","sign_citation":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/action/citation_signature","submit_replication":"https://pith.science/pith/Z4CGJHK5XDYLLS7QSLH4IGGQ7F/action/replication_record"}},"created_at":"2026-07-05T10:24:45.468883+00:00","updated_at":"2026-07-05T10:24:45.468883+00:00"}