Showing cs.LGShow all
2 papers · 1 filter
cs.LG2026
Modular Norm RandOpt: Population-Efficient Ensembling through Architecture-Aware Perturbations
Kirato Yoshihara, Hiroaki Hamade
RandOpt samples weight-perturbed language models and ensembles top-ranked candidates through plurality voting, but its global perturbation scale ignores heterogeneous module geomet…
cs.LG2026
Different Layers, Different Manifolds: Module-Wise Weight-Space Geometry in Transformer Optimization
Kirato Yoshihara
Weight-space geometry plays a central role in neural network optimization, yet manifold constraints are often applied uniformly across all weight matrices. In this work, we ask whe…