Beyond Linearity in Attention Projections: The Case for Nonlinear Queries

Marko Karbevski

Track: tiny paper (up to 4 pages)

Keywords: MLP, Attention, Redundancy, Non-linearity, Invariance, Transformers

TL;DR: Motivated by the theoretical redundancy of the linear Query projection, we replace it with a nonlinear residual MLP and demonstrate improved performance on GPT-3 small scale models

Abstract: Recent algebraic analysis shows that in decoder-only and encoder-only transformers, the Query projection $W_Q$ may be set to identity without noticeable performance deterioration. This is possible because attention depends on $X$ only through the products $XW_Q, XW_K, XW_V$, allowing basis transformations to be absorbed by adjacent layers and propagated through the network. We replace $W_Q \in \mathbb{R}^{d \times d}$ with a nonlinear residual of the form $Q(X) = X + f_\theta(X)$, where $f_\theta$ is a bottleneck MLP with $d^2 + O(d)$ parameters. The identity term anchors the nonlinearity to a known-good prior. Experiments on GPT-3 small style models show consistent improvement over the baseline, comfortably outperforming a model with 12.5\% more non-embedding parameters. These results motivate investigation at larger scales and across modalities.

Anonymization: This submission has been anonymized for double-blind review via the removal of identifying information such as names, affiliations, and identifying URLs.

Submission Number: 82