Since its establishment in January 2019 with funding from the Ministry of Science and Technology (now the National Science and Technology Council, NSTC), the NTU-IBM Quantum Computing Center has continuously provided a quantum computing platform for academic research and interdisciplinary collaboration in Taiwan, while also taking on the important mission of promoting quantum computing education nationwide. Since the first Quantum Computing Hackathon was held in 2020 at the NTU Xitou Education Center, the event was suspended in 2021 due to the pandemic, but was subsequently held annually at Syntrend Creative Park 11F in Taipei in 2022, 2023, 2024, and 2025. From August 12 to 14, 2026, the IBM Quantum User Meeting and the three-day 2026 Quantum Computing Qiskit Hackathon were held on 1F, Department of Physics and Center for Condensed Matter Sciences, NTU. (Event website: https://quantum.ntu.edu.tw/)
This year’s Hackathon attracted over 80 registrants, the majority of whom already had experience with IBM Quantum systems, showcasing the significant progress made in promoting quantum computing in recent years. In addition to local participants, the event continued to welcome international partners, including Keio University (Japan), Yonsei University (Korea), and Czech Technical University in Prague (Czech Republic), each sending 1–6 students to join, making the competition more international, diverse, and challenging.
The opening ceremony featured remarks by NSTC Department of Natural Sciences and Sustainable Development General Director Hung-Wen Li and NSTC Quantum System Promotion Working Group Convener Prof. Wen-Hao Chang, followed by a keynote lecture delivered by Mr. Takeshi Watanabe of IBM Quantum, who shared the latest trends in quantum computing. To enhance the challenge and excitement, this year’s Hackathon once again revealed the competition topics on-site, with team formation taking place during the event. Participants were required to design quantum programs, obtain results using Qiskit, and deliver oral presentations in English, evaluated by both peers and judges. Evaluation criteria included: Originality and Uniqueness, Usefulness and Complexity, Potential Quantum Community Benefit, and Presentation and Team Collaboration. Winning teams were selected by the jury, bringing the 2026 Hackathon to a successful conclusion.
This year’s event was primarily funded by the NSTC and supported by IBM Quantum,which provided on-site coaches to guide participants. Co-organizers included the Center for Quantum Science and Engineering at NTU, the Center for Advanced Computing and Imaging in Biomedicine at NTU, the National Center for Theoretical Sciences – Physics Division, the Taiwan Association of Quantum Computation and Information Technology (TAQCIT), and the Chung Yuan Christian University Quantum Information Center. Domestic companies also actively supported the event, not only by sending employees to participate in the competition but also by sponsoring special awards. In addition to the NSTC Jury Prize and the NTU-IBM Quantum Computing Center Prize, this year’s Enterprise Special Awards were sponsored by Foxconn, Formosa Plastics Corporation, Wistron, and Keysight Technologies.
Cultivating cross-disciplinary talents with practical experience, as well as developing essential quantum applications, forms a critical foundation for Taiwan’s long-term quantum technology development. Promoting quantum computing education and widespread learning is a key driver in advancing Taiwan’s quantum industry to the next level.
The following is a summary of the competition reports submitted by each winning team:
Group 1 developed a Variational Quantum Circuit (VQC)-based deep reinforcement learning framework using Double Deep Q-Networks (DQNs), experience replay, ε-greedy exploration, reward shaping, and 3-step returns for the CartPole-v1 control task. Their quantum model employs a five-layer data-reuploading variational circuit with only 70 trainable parameters, substantially fewer than the 1282 parameters of the standard classical Multilayer Perceptron (MLP) baseline. The team’s contributions include testing on the IBM Marrakesh real quantum hardware, validating the fast quantum executor to an accuracy of 10−7, and open-sourcing the complete codebase and tutorial resources. Awarded the Enterprise Special Prize.
Group 11 investigated how Sample-Based Quantum Diagonalization (SQD) can simulate the Fermi-Hubbard model using Variational Quantum Algorithm (VQA)-based quantum circuits, from ground-state estimation to local electron-removal responses. Using a 6-site periodic Hubbard ring, the team implemented VQA–SQD workflows with error-suppression techniques on IBM Quantum hardware. They indicated that SQD can improve ground-state energies and reproduce electron-removal spectral features, particularly for localized systems. The team also implemented the non-unitary electron-removal operation using an ancilla qubit and post-selection, finding good agreement between direct SQD and quantum-circuit responses. Their work highlights both the potential and limitations of SQD. Awarded the Enterprise Special Prize.
Group 14 developed a reproducible hybrid quantum reinforcement learning framework to investigate the effectiveness of Variational Quantum Circuits (VQCs) on reinforcement learning tasks. Using CartPole-v1 as the primary benchmark, the team implemented and compared several quantum approaches, including Quantum Deep Q-Networks (QDQNs), Quantum Policy Gradient (QPG), and hybrid Q2C / A2Q architectures, against parameter-matched classical Multilayer Perceptron (MLP) baselines. Their results showed that QDQN achieved faster training than the other approaches, solving 3/5 trials with only 46 trainable parameters. Additionally, they successfully reproduced all 16/16 action decisions on the real hardware IBM_Marrakesh. The team also introduced Spinpole, a new quantum-control game, demonstrating broader applications of quantum reinforcement learning. Awarded the Enterprise Special Prize.
Group 13 proposed applying Approximate Quantum Compilation with Tensor Networks (AQC-Tensor) to each Krylov-power circuit within Sampling-based Krylov Quantum Diagonalization (SKQD). This compresses the time-evolution circuits, which would otherwise grow deeper as the Krylov dimension increases, while preserving SKQD’s convergence guarantee to the ground state. The result is a significant reduction in the circuit depth required on hardware, making the method better suited for scaling on current noisy quantum devices. Awarded the Enterprise Special Prize.
Group 4 focused on overcoming catastrophic forgetting through Quantum Continual Learning (QCL) — the problem where training a model on a new task significantly degrades its performance on previously learned tasks when tasks are learned sequentially. The team built a variational quantum classifier and trained it sequentially across multiple tasks, evaluating the degree of forgetting. They implemented two methods, Measurement-based Parameter Isolation (MPI) and Quantum Generative Reply (QGR), to improve knowledge retention, and validated their approach on real QPU hardware. Awarded the NTU-IBM Quantum Hub Prize.
Group 10 used the Quantum Approximate Optimization Algorithm (QAOA) to simulate the frustrated J1–J2 Ising Hamiltonian on a square lattice. Because the nearest-neighbor (J1) and next-nearest-neighbor (J2) interactions compete, the ground-state configuration is nontrivial to determine. The team constructed the corresponding Ising Hamiltonian and used QAOA to find bitstrings representing the ground state, from which they estimated the ground-state energy, using a classical simulation as a benchmark for comparison. They implemented the fundamental version on a simulator with a 3×3 (9-qubit) lattice and subsequently extended the experiment to a 6×6 lattice on real IBM hardware (Heron/Nighthawk). Awarded the Grand Jury Prize.
