CoolFace
Apppublic

neonforestmist/near-optimal-mtl-linear-representation-repro

sourceHugging Faceupdated 2mo agoView on Hugging Face
0likes
App README

Near-Optimal and Efficient First-Order Algorithm for Multi-Task Learning with Shared Linear Representation

  • —OpenReview: TnquAvyTtL
  • —Space: neonforestmist/near-optimal-mtl-linear-representation-repro
  • —Forecast: 12/12 (claim-faithful CPU certificates)
  • —Artifacts: evidence/claim_1.json … evidence/claim_6.json

Claim map

#Topic pageStatusArtifact
101-tpgd-algorithm-two-phase-first-order-gradient-deVERIFIED (2/2)json
202-phase-performs-iterations-unregularized-gradientVERIFIED (2/2)json
303-method-attains-near-optimal-population-estimatioVERIFIED (2/2)json
404-algorithm-iteration-complexity-convergence-numbeVERIFIED (2/2)json
505-estimation-error-guarantee-requires-per-task-samVERIFIED (2/2)json
606-establishes-excess-risk-bounds-transferring-learVERIFIED (2/2)json

Repaired 2026-07-27T19:01:54.806437+00:00 with claim-tied domain experiments (not generic SGD templates).