rdubwiley/near-optimal-and-efficient-first-order-algorithm-repro
0
1{2 "schema_version": 1,3 "title": "Repro - Near-Optimal and Efficient First-Order Algorithm for Multi-Task Learning with Shared Linear Representation",4 "emoji": "🎯",5 "space_id": "rdubwiley/near-optimal-and-efficient-first-order-algorithm-repro",6 "paper": {7 "arxiv_id": "2605.00473"8 },9 "tags": [10 "icml2026-repro",11 "paper-TnquAvyTtL"12 ],13 "updated_at": "2026-07-19T02:53:04+00:00",14 "root": {15 "slug": "index",16 "title": "Repro - Near-Optimal and Efficient First-Order Algorithm for Multi-Task Learning with Shared Linear Representation",17 "file": "pages/index.md",18 "children": [19 {20 "slug": "claim-1-first-order-algorithm-for-multi-task-learning-with-shared-linear-representation-converges-in-1-iterations",21 "title": "Claim 1: First-order algorithm for multi-task learning with shared linear representation converges in Õ(1) iterations",22 "file": "pages/claim-1-first-order-algorithm-for-multi-task-learning-with-shared-linear-representation-converges-in-1-iterations/page.md",23 "children": []24 },25 {26 "slug": "claim-2-achieves-near-optimal-estimation-error-of-dk-tn-improving-over-existing-methods-by-factor-k",27 "title": "Claim 2: Achieves near-optimal estimation error of Õ(dk/(TN)) improving over existing methods by factor k",28 "file": "pages/claim-2-achieves-near-optimal-estimation-error-of-dk-tn-improving-over-existing-methods-by-factor-k/page.md",29 "children": []30 },31 {32 "slug": "claim-3-guarantees-computational-efficiency-for-jointly-learning-shared-representation-and-task-specific-parameters",33 "title": "Claim 3: Guarantees computational efficiency for jointly learning shared representation and task-specific parameters",34 "file": "pages/claim-3-guarantees-computational-efficiency-for-jointly-learning-shared-representation-and-task-specific-parameters/page.md",35 "children": []36 },37 {38 "slug": "conclusion",39 "title": "Conclusion",40 "file": "pages/conclusion/page.md",41 "children": []42 }43 ]44 },45 "agent_view_tokens": 6709,46 "revision": "1784429584702297171"47}