# Twirlator: A Pipeline for Analyzing Subgroup Symmetry Effects in Quantum Machine Learning Ansatzes

arXiv: https://arxiv.org/abs/2511.04243 · HTML: https://mlsanity.com/paper/2511.04243/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 27 references that are in the map.
- First posted on arXiv 6 Nov 2025.

## Builds on (most cited first, 10 of 10)

- [Semi-Supervised Classification with Graph Convolutional Networks](https://mlsanity.com/paper/1609.02907/) (2016)
- [Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges](https://mlsanity.com/paper/2104.13478/) (2021)
- [Group Equivariant Convolutional Networks](https://mlsanity.com/paper/1602.07576/) (2016)
- [E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials](https://mlsanity.com/paper/2101.03164/) (2021)
- [Barren plateaus in quantum neural network training landscapes](https://mlsanity.com/paper/1803.11173/) (2018)
- [Residual Pathway Priors for Soft Equivariance Constraints](https://mlsanity.com/paper/2112.01388/) (2021)
- [Theoretical Guarantees for Permutation-Equivariant Quantum Neural Networks](https://mlsanity.com/paper/2210.09974/) (2022)
- [Agent-Q: Fine-Tuning Large Language Models for Quantum Circuit Generation and Optimization](https://mlsanity.com/paper/2504.11109/) (2025)
- [High-performance training and inference for deep equivariant interatomic potentials](https://mlsanity.com/paper/2504.16068/) (2025)
- [Symmetry-invariant quantum machine learning force fields](https://mlsanity.com/paper/2311.11362/) (2023)

## Abstract (arXiv)

Symmetry is a strong inductive bias in geometric deep learning and its quantum counterpart, and has attracted increasing attention for improving the trainability of QML models. Yet incorporating symmetries into quantum machine learning (QML) ansatzes is not free: symmetrization often adds gates and constrains the circuits. To understand these effects, we present Twirlator, which is an automated pipeline that symmetrizes parameterized QML ansatzes and quantifies the trade-offs as the amount of symmetry increases. Twirlator models partial symmetries by the size of a subgroup of the symmetric group, enabling analysis between the ``no symmetry'' and ``full symmetry'' extremes. Across 19 common ansatz patterns, Twirlator symmetrizes circuits with respect to any subgroup of Sn and measures (1) generator drift, (2) circuit overhead (depth and size), and (3) expressibility and entangling capability. The experimental evaluation focuses on subgroups of S4 and S5. Twirlator reveals that larger subgroups typically increase circuit overhead, reduce expressibility, and often increase entangling capability. The pipeline and results provide practical guidance for selecting ansatz patterns and symmetry levels that balance hardware cost and model performance in symmetry-aware QML applications.