Student research group / Wrocław University of Science and Technology

Quantum machine learning, measured on the processor.

We design variational circuits for the native gate set of ODRA 5, Poland's first superconducting quantum computer, and report what they do on the hardware itself rather than only in simulation.

Our Research Join Us

Current state of the programme

Hardware: ODRA 5 / IQM Spark / 5 qubits
Presented: Oral talk, IEEE qCCL 2026
In preparation: Two AAAI-27 submissions

Read the flagship study →

[01]

Mission

Est. 2025 / Wrocław

Simulator results are a hypothesis. The processor is the experiment.

Axion Research is a student research group at Wrocław University of Science and Technology. Our main line of work is quantum machine learning: designing and benchmarking variational circuits on ODRA 5, Poland's first superconducting quantum computer. Alongside it we work on machine learning systems.

We present our work. Our first peer-reviewed abstract was accepted for an oral talk at an IEEE conference in 2026, and further work is in preparation for AAAI-27. The experiments run on a physical quantum processor rather than only simulators, and we treat reproducibility as part of the result: shared library code, versioned cross-validation folds, and checkpointed weights.

Members: 12
Conference Contributions: 2
Founded: 2025

[02]

Research

Quantum ML / ML Systems

Quantum Machine Learning

Variational circuits designed for the native gate set of a real processor, then measured on it: compiled cost, fidelity, expressibility, entanglement, and end-to-end accuracy on ODRA 5.

Presented as an oral talk at IEEE qCCL 2026

Machine Learning Systems

How models behave once the cost of running them matters: inference efficiency, measurement methodology, and applied modelling on real datasets.

AAAI-27 submission in preparation

Earlier work: EEG and brain-computer interfaces, most recently a poster at MCSB 2026.

[03]

Featured Research

View All →

Flagship Research Project

QC1: Quantum Banknote Classifier on ODRA 5

Axion's flagship quantum research program: a hardware-aware variational quantum classifier for banknote authentication on ODRA 5, comparing IQM Spark-native ansatze against simulator-oriented designs.

5-qubit register

Head to head: classification on the same folds and protocol, compilation at the same optimization level

ODRA-native Simulator-oriented

Accuracy on the physical QPU, depth 4, five-fold mean 87.6% 74.5%

Compiled circuit depth, depth 6, optimization level 1 127 263

Native two-qubit gates, depth 6, optimization level 1 60 87

Hack4SAGES 2026 · 10 March 2026 - 12 March 2026

ExoBiome

A 3-day Hack4SAGES 2026 project building a hybrid quantum-classical regression stack for exoplanet transmission spectra to estimate H2O, CO2, CO, CH4, and NH3.

Python PyTorch Quantum ML

View Project →

Pipeline

In the Works

Two AAAI-27 submissions are in preparation, and the ODRA 5 study is being extended beyond its first benchmark.

See what is next →

qCCL 2026 conference group photo, Aalborg

Conference / July 2026

We presented QC1 at IEEE qCCL 2026 in Aalborg.

Talks on quantum control, computing, and learning, with our own talk on the programme: hardware-efficient ansatz design for IQM Spark. For several of us it was a first talk at a scientific conference, in the first year of a bachelor's degree. Quantum processor access came from the Wrocław Centre for Networking and Supercomputing, and the work was carried out at the Department of Artificial Intelligence.

Read the study →