{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:SOVHKOTWL33UZ2SJBZ5ZX3BN3L","short_pith_number":"pith:SOVHKOTW","schema_version":"1.0","canonical_sha256":"93aa753a765ef74cea490e7b9bec2ddaf1425bad6220e76045ef026ea2b6f7a8","source":{"kind":"arxiv","id":"2103.02631","version":3},"attestation_state":"computed","paper":{"title":"RotoGrad: Gradient Homogenization in Multitask Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Adri\\'an Javaloy, Isabel Valera","submitted_at":"2021-03-03T19:03:52Z","abstract_excerpt":"Multitask learning is being increasingly adopted in applications domains like computer vision and reinforcement learning. However, optimally exploiting its advantages remains a major challenge due to the effect of negative transfer. Previous works have tracked down this issue to the disparities in gradient magnitudes and directions across tasks, when optimizing the shared network parameters. While recent work has acknowledged that negative transfer is a two-fold problem, existing approaches fall short as they only focus on either homogenizing the gradient magnitude across tasks; or greedily ch"},"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":"2103.02631","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-03-03T19:03:52Z","cross_cats_sorted":[],"title_canon_sha256":"352acf6872192281fa61d82c3d9eb1344d86bfc81a74b050289a63c819ab0aa1","abstract_canon_sha256":"666c6b5c5bd5d3282f3c3b97cbe38b11c48411811f7969f4766ddb0decdf406a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:57:25.967005Z","signature_b64":"60VPa+ce6ijBHG0iwH80pwp5kjmM+wfL6whap15S9APpNmUUWzscRw6ihShiKoMIoApdHTqzWnCekR1i/XDYCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"93aa753a765ef74cea490e7b9bec2ddaf1425bad6220e76045ef026ea2b6f7a8","last_reissued_at":"2026-07-05T03:57:25.966500Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:57:25.966500Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RotoGrad: Gradient Homogenization in Multitask Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Adri\\'an Javaloy, Isabel Valera","submitted_at":"2021-03-03T19:03:52Z","abstract_excerpt":"Multitask learning is being increasingly adopted in applications domains like computer vision and reinforcement learning. However, optimally exploiting its advantages remains a major challenge due to the effect of negative transfer. Previous works have tracked down this issue to the disparities in gradient magnitudes and directions across tasks, when optimizing the shared network parameters. While recent work has acknowledged that negative transfer is a two-fold problem, existing approaches fall short as they only focus on either homogenizing the gradient magnitude across tasks; or greedily ch"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2103.02631","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/2103.02631/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":"2103.02631","created_at":"2026-07-05T03:57:25.966566+00:00"},{"alias_kind":"arxiv_version","alias_value":"2103.02631v3","created_at":"2026-07-05T03:57:25.966566+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2103.02631","created_at":"2026-07-05T03:57:25.966566+00:00"},{"alias_kind":"pith_short_12","alias_value":"SOVHKOTWL33U","created_at":"2026-07-05T03:57:25.966566+00:00"},{"alias_kind":"pith_short_16","alias_value":"SOVHKOTWL33UZ2SJ","created_at":"2026-07-05T03:57:25.966566+00:00"},{"alias_kind":"pith_short_8","alias_value":"SOVHKOTW","created_at":"2026-07-05T03:57:25.966566+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27377","citing_title":"DanceOPD: On-Policy Generative Field Distillation","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30370","citing_title":"MUSE: Unlocking Timestep as Native Task Steering for One-Step Dense Prediction","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05660","citing_title":"Distributionally Robust Multi-Objective Optimization","ref_index":49,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L","json":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L.json","graph_json":"https://pith.science/api/pith-number/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/graph.json","events_json":"https://pith.science/api/pith-number/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/events.json","paper":"https://pith.science/paper/SOVHKOTW"},"agent_actions":{"view_html":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L","download_json":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L.json","view_paper":"https://pith.science/paper/SOVHKOTW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2103.02631&json=true","fetch_graph":"https://pith.science/api/pith-number/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/graph.json","fetch_events":"https://pith.science/api/pith-number/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/action/storage_attestation","attest_author":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/action/author_attestation","sign_citation":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/action/citation_signature","submit_replication":"https://pith.science/pith/SOVHKOTWL33UZ2SJBZ5ZX3BN3L/action/replication_record"}},"created_at":"2026-07-05T03:57:25.966566+00:00","updated_at":"2026-07-05T03:57:25.966566+00:00"}