{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:XQXBCMC54PJZKVIV4FNUNCPQ2R","short_pith_number":"pith:XQXBCMC5","schema_version":"1.0","canonical_sha256":"bc2e11305de3d3955515e15b4689f0d47adec483b3a8011a07aadb1b1ffc2337","source":{"kind":"arxiv","id":"2005.00944","version":1},"attestation_state":"computed","paper":{"title":"Understanding and Improving Information Transfer in Multi-Task Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Christopher R\\'e, Hongyang R. Zhang, Sen Wu","submitted_at":"2020-05-02T23:43:52Z","abstract_excerpt":"We investigate multi-task learning approaches that use a shared feature representation for all tasks. To better understand the transfer of task information, we study an architecture with a shared module for all tasks and a separate output module for each task. We study the theory of this setting on linear and ReLU-activated models. Our key observation is that whether or not tasks' data are well-aligned can significantly affect the performance of multi-task learning. We show that misalignment between task data can cause negative transfer (or hurt performance) and provide sufficient conditions f"},"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":"2005.00944","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-05-02T23:43:52Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"73551a46a297cb12ed178cd62f386a538c7895b820224980f499d3e69bad02be","abstract_canon_sha256":"16a743ba15f7bfbad68f508a556e89f9e4670724294fb94e6e37d892dac0a5ee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:00:02.416777Z","signature_b64":"wNX3yp1W/bZCeRDgkv9/8k5JhEU+aqnI8MiddJDsOb2DuJMAsaVWu0151ieUHUZ1ZT2rrb3YS1JeEhT4lcYsBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc2e11305de3d3955515e15b4689f0d47adec483b3a8011a07aadb1b1ffc2337","last_reissued_at":"2026-07-05T01:00:02.416377Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:00:02.416377Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding and Improving Information Transfer in Multi-Task Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Christopher R\\'e, Hongyang R. Zhang, Sen Wu","submitted_at":"2020-05-02T23:43:52Z","abstract_excerpt":"We investigate multi-task learning approaches that use a shared feature representation for all tasks. To better understand the transfer of task information, we study an architecture with a shared module for all tasks and a separate output module for each task. We study the theory of this setting on linear and ReLU-activated models. Our key observation is that whether or not tasks' data are well-aligned can significantly affect the performance of multi-task learning. We show that misalignment between task data can cause negative transfer (or hurt performance) and provide sufficient conditions f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2005.00944","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/2005.00944/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":"2005.00944","created_at":"2026-07-05T01:00:02.416440+00:00"},{"alias_kind":"arxiv_version","alias_value":"2005.00944v1","created_at":"2026-07-05T01:00:02.416440+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2005.00944","created_at":"2026-07-05T01:00:02.416440+00:00"},{"alias_kind":"pith_short_12","alias_value":"XQXBCMC54PJZ","created_at":"2026-07-05T01:00:02.416440+00:00"},{"alias_kind":"pith_short_16","alias_value":"XQXBCMC54PJZKVIV","created_at":"2026-07-05T01:00:02.416440+00:00"},{"alias_kind":"pith_short_8","alias_value":"XQXBCMC5","created_at":"2026-07-05T01:00:02.416440+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24259","citing_title":"SURGELLM: Rethinking Multi-Task Evaluation through Task-Aware Feature Gating with Class-Balanced Normalization","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30615","citing_title":"Tuning-Free Efficient Estimation for Multi-Source Data via Covariance-Aware Shrinkage","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17126","citing_title":"Multi-task Linear Regression without Eigenvalue Lower Bounds: Adaptivity, Robustness, and Safety","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19301","citing_title":"iGSP:Implicit Gradient Subspace Projection for Efficient Continual Learning of Vision-Language Models","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2507.16433","citing_title":"Adaptive Multi-task Learning for Multi-sector Portfolio Optimization","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R","json":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R.json","graph_json":"https://pith.science/api/pith-number/XQXBCMC54PJZKVIV4FNUNCPQ2R/graph.json","events_json":"https://pith.science/api/pith-number/XQXBCMC54PJZKVIV4FNUNCPQ2R/events.json","paper":"https://pith.science/paper/XQXBCMC5"},"agent_actions":{"view_html":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R","download_json":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R.json","view_paper":"https://pith.science/paper/XQXBCMC5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2005.00944&json=true","fetch_graph":"https://pith.science/api/pith-number/XQXBCMC54PJZKVIV4FNUNCPQ2R/graph.json","fetch_events":"https://pith.science/api/pith-number/XQXBCMC54PJZKVIV4FNUNCPQ2R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R/action/storage_attestation","attest_author":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R/action/author_attestation","sign_citation":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R/action/citation_signature","submit_replication":"https://pith.science/pith/XQXBCMC54PJZKVIV4FNUNCPQ2R/action/replication_record"}},"created_at":"2026-07-05T01:00:02.416440+00:00","updated_at":"2026-07-05T01:00:02.416440+00:00"}