Equivariance is a fundamental property in computer vision models, yet strict equivariance is rarely satisfied in real-world data, which can limit a model's performance. Controlling the degree of equivariance is therefore desirable. We propose a general framework for constructing soft equivariant models by projecting the model weights into a designed subspace. The method applies to any pre-trained architecture and provides theoretical bounds on the induced equivariance error. Empirically, we demonstrate the effectiveness of our method on multiple pre-trained backbones, including ViT and ResNet, across image classification, semantic segmentation, and human-trajectory prediction tasks. Notably, our approach improves the performance while simultaneously reducing equivariance error on the competitive ImageNet benchmark.
@article{arxiv.2603.26657,
title = {Tunable Soft Equivariance with Guarantees},
author = {Md Ashiqur Rahman and Lim Jun Hao and Jeremiah Jiang and Teck-Yian Lim and Raymond A. Yeh},
journal= {arXiv preprint arXiv:2603.26657},
year = {2026}
}