AWhat changed
Assigned a separate input-dependent gating network to each task while sharing expert subnetworks, letting different prediction towers use different mixtures of learned representations.
Scientific scope
Assigned a separate input-dependent gating network to each task while sharing expert subnetworks, letting different prediction towers use different mixtures of learned representations. Extends mixture-of-experts for task-relationship modeling; it does not invent MoE or sparse conditional execution. Softmax mixtures need not select only one expert. Reduced negative transfer and trainability gains are empirical for studied task correlations, not guarantees for all task pairs.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Extends mixture-of-experts for task-relationship modeling; it does not invent MoE or sparse conditional execution. Softmax mixtures need not select only one expert. Reduced negative transfer and trainability gains are empirical for studied task correlations, not guarantees for all task pairs.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction