# Group Equivariant Stand-Alone Self-Attention For Vision

Source: [Group Equivariant Stand-Alone Self-Attention For Vision](<https://www.sota2.com/research/paper/68fa69961e209131cea22e0c>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-27T06:06:39.448Z

Authors: David W. Romero, Jean-Baptiste Cordonnier

Published: 2020-10-02T13:16:00.000Z

## Abstract

We provide a general self-attention formulation to impose group equivariance to arbitrary symmetry groups. This is achieved by defining positional encodings that are invariant to the action of the group considered. Since the group acts on the positional encoding directly, group equivariant self-attention networks (GSA-Nets) are steerable by nature. Our experiments on vision benchmarks demonstrate consistent improvements of GSA-Nets over non-equivariant self-attention networks.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Image Classification on MNIST Rotated](<https://www.sota2.com/research/sota/image-classification-on-mnist-rotated.md>): Image Classification; MNIST Rotated; Accuracy (%): 97.9.

- [Image Classification on PatchCamelyon](<https://www.sota2.com/research/sota/image-classification-on-patchcamelyon.md>): Image Classification; PatchCamelyon; Accuracy: 84.76.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2010.00977>)

[PDF](<https://arxiv.org/pdf/2010.00977.pdf>)