Publications

→

Preprints

(2)

Modular connectivity in neural networks emerges from Poisson noise-motivated regularisation, and promotes robustness and compositional generalisation

Daoyuan Qian, Qiyao Liang, Ila Fiete

arXiv preprint · 2025

Under review at Physical Review X Life, by invitation. We show how modular connectivity in neural networks emerges from Poisson noise-motivated regularisation, and how it promotes robustness and compositional generalisation.

Read

Evolution Strategies at Scale: LLM Fine-Tuning Beyond Reinforcement Learning

Xin Qiu*, Yulu Gan*, Conor F. Hayes*, Qiyao Liang, Elliot Meyerson, Babak Hodjat, Risto Miikkulainen

arXiv preprint · 2025

Under review. We explore evolution strategies at scale for LLM fine-tuning as an alternative to reinforcement learning approaches.

Read

→

Conference Proceedings

(2)

Compositional Generalization via Forced Rendering of Disentangled Latents

Qiyao Liang*, Daoyuan Qian*, Liu Ziyin, Ila Fiete

ICML 2025 · 2025

Standard networks given separate x/y cues still memorize and stitch examples rather than combine them. Forcing each cue into the output space—via low-rank embeddings or simple stripe data—enables true compositional generalization.

Read

How Diffusion Models Learn to Factorize and Compose

Qiyao Liang, Ziming Liu, Mitchell Ostrow, Ila R. Fiete

NeurIPS 2024 · 2024

Diffusion models are capable of generating photo-realistic images that combine elements which likely do not appear together in the training set, demonstrating the ability to compositionally generalize. We explore how diffusion models factorize and compose concepts through controlled experiments on 2D Gaussian data.

Read

→

Journal Publications

(5)

Pulse optimization for high-precision motional-mode characterization in trapped-ion quantum computers

Qiyao Liang, Mingyu Kang, Ming Li, Yunseong Nam

Quantum Science Technology 9 035007 · 2024

High-fidelity operation of quantum computers requires precise knowledge of the physical system through characterization. We develop an alternative scheme that leverages the degrees of freedom in pulse design to actively silence the effects of cross-mode coupling.

Read

Modeling the Performance of Early Fault-Tolerant Quantum Algorithms

Qiyao Liang, Yiqing Zhou, Archismita Dalal, Peter D. Johnson

Physical Review Research 6 023118 · 2023

Progress in fault-tolerant quantum computation has driven the pursuit of practical applications with early fault-tolerant quantum computers. We propose a methodology for modeling algorithm performance on EFTQC devices under varying degrees of error.

Read

Efficient motional-mode characterization for high-fidelity trapped-ion quantum computing

Mingyu Kang, Qiyao Liang, Ming Li, Yunseong Nam

Quantum Science Technology 8 024002 · 2023

To achieve high-fidelity operations on a large-scale quantum computer, the parameters of the physical system must be efficiently characterized with high accuracy. We develop and explore physical models that accurately predict both magnitude and sign of the Lamb–Dicke parameters.

Read

Batch Optimization of Frequency-Modulated Pulses for Robust Two-Qubit Gates in Ion Chains

Mingyu Kang, Qiyao Liang, Bichen Zhang, Shilin Huang, Ye Wang, Chao Fang, Jungsang Kim, Kenneth R. Brown

Physical Review Applied 16 024039 · 2021

Two-qubit gates in trapped-ion quantum computers are generated by applying spin-dependent forces. We improve the robustness of frequency-modulated Mølmer-Sørensen gates to motional mode-frequency offsets by optimizing the average performance over a range of systematic errors.

Read

High-Fidelity Two-Qubit Gates Using a Microelectromechanical-System-Based Beam Steering System for Individual Qubit Addressing

Ye Wang, Stephen Crain, Chao Fang, Bichen Zhang, Shilin Huang, Qiyao Liang, Pak Hong Leung, Kenneth R. Brown, Jungsang Kim

Physical Review Letter 125 150505 · 2020

We realize and characterize high-fidelity two-qubit gates in a system with up to four ions using radial modes. The ions are individually addressed by two tightly focused beams steered using microelectromechanical system mirrors.

Read

→

Workshops

(1)

Do Diffusion Models Learn Semantically Meaningful and Efficient Representations?

Qiyao Liang, Ziming Liu, Ila R. Fiete

ICLR 2024 Workshop ME-FoMo · 2024

We perform controlled experiments on conditional DDPMs learning to generate 2D spherical Gaussian bumps. Our results show that the emergence of semantically meaningful latent representations is key to achieving high performance.

Read