When Does Dense Retrieval Need Asymmetric Geometry? A Bias-Variance Theory of Shared and Dual Projections
A bias-variance theory for low-rank bilinear scoring is introduced and a local Gaussian boundary is proved: dual has lower risk exactly when squared directional signal exceeds the estimation cost of its additional degrees of freedom.
Maojun Sun, Yan-Cheng Yuan, Jian Huang et al.
· 0 citations