paper-with-me

Papers

Beyond Linearity in Attention Projections: The Case for Nonlinear Queries

2026-03-11 · Marko Karbevski arxiv

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 $W_Q \in \R^{d \times d}$ with a nonlinear residual of the form $Q(X) = X + f_θ(X)$, where $f_θ$ 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 ($2.40\%$ lower validation log-loss, $6.81\%$ lower perplexity), comfortably outperforming a model with 12.5\% more non-embedding parameters. These results motivate investigation at larger scales and across modalities.

📄 PDF Abstract BibTeX arXiv:2603.13381

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Controlling Grokking with Nonlinearity and Data Symmetry

2024-11-08 · Ahmed Salah, David Yevick

This paper demonstrates that grokking behavior in modular arithmetic with a modulus P in a neural network can be controlled by modifying the profile of the activation function as well as the depth and width of the model.…

Sliced ReLU attention: Quasi-linear contextual expressivity via sorting

2025-12-12 · François-Xavier Vialard, Siwan Boufadène arxiv

We introduce sliced ReLU attention, a new attention mechanism that departs structurally from both softmax and its approximation alternatives. Instead of applying a nonlinearity to pairwise dot products, we operate on one…

Rethinking Nonlinearity: Trainable Gaussian Mixture Modules for Modern Neural Architectures

2025-10-08 · Weiguo Lu, Gangnan Yuan, Hong-kun Zhang, Shangyang Li arxiv

Neural networks in general, from MLPs and CNNs to attention-based Transformers, are constructed from layers of linear combinations followed by nonlinear operations such as ReLU, Sigmoid, or Softmax. Despite their strengt…

Zero-Shot Audio Classification with Factored Linear and Nonlinear Acoustic-Semantic Projections

2020-11-25 · Huang Xie, Okko Räsänen, Tuomas Virtanen

In this paper, we study zero-shot learning in audio classification through factored linear and nonlinear acoustic-semantic projections between audio instances and sound classes. Zero-shot learning in audio classification…

Audio ClassificationClassificationGeneral ClassificationZero-shot Audio Classification+1

Nexusformer: Nonlinear Attention Expansion for Stable and Inheritable Transformer Scaling

2026-04-21 · Weijie Zhao, Mingquan Liu, Bolun Wang, Simo Wu 외 arxiv

Scaling Transformers typically necessitates training larger models from scratch, as standard architectures struggle to expand without discarding learned representations. We identify the primary bottleneck in the attentio…