The 2nd Quantum AI Competition

1. Overview

The 2nd Quantum AI Online Challenge is a hands-on hackathon where participants design quantum machine learning (QML) models, submit their quantum circuits as JSON, and see their performance evaluated on a hidden test set using a GPU-based grading system. Participants start from tutorials, experiment with quantum circuits in simulation, and compete on a public leaderboard. Finalists will have the opportunity to run their models on a real quantum device (details to be announced).

2. Theme

“From Quantum Circuits to Real-World Inference”
Design, implement, and evaluate a quantum model that performs classification on a given dataset, under fixed resource constraints (qubit count, circuit depth, measurement scheme, etc.).

Both non-developers and AI/ML developers are welcome. Official tutorials will guide complete beginners from basic concepts to submitting a working model.

3. Organizers & Sponsors

Hosts / Organizers
- NORMA
- AIFactory
Sponsors
- TBD

4. Important Dates (TBD, to be finalized)

Exact dates will be announced on the official AIFactory challenge page. The overall structure will be:

Registration Period: TBD
Team Building Deadline: TBD
Online Preliminary Round (simulation on AIFactory): TBD
Announcement of Finalists: TBD
On-site Final Round (Jeonju University): TBD
Award Ceremony: TBD

5. Eligibility & Team Formation

Who can join
- AI/ML developers, software engineers, data scientists
- Students and non-CS majors with an interest in quantum computing
Team size: 1–5 members per team
Participants must agree to the competition rules and code of conduct.

6. How to Participate

Click [Join] on the challenge page and complete the registration form.
Form or join a team (up to 5 members).
Complete the pre-challenge tutorials:
- Basics of quantum circuits and the chosen framework (e.g., PennyLane + PyTorch)
- How to encode data into quantum circuits
- How to prepare the final JSON submission
Download the starter kit:
- Example Jupyter notebooks
- Environment setup scripts
- Sample submission JSON file
Design and train your quantum model in your local or cloud environment.
Generate the required JSON submission file and upload it to the AIFactory platform.
Monitor your score and visual diagnostics on the leaderboard page.

7. Tutorials & Learning Materials

Before and during the preliminary round, the following tutorial track will be provided (online):

Quantum Circuits & PennyLane Basics
- Qubits, gates, and measurement
- Defining and visualizing quantum circuits
- Choosing backends and basic measurement (expval, probs, etc.)
QML & Torch Interface
- Parameterized Quantum Circuits (PQC)
- Batched parameters
- Integrating quantum circuits with PyTorch
Quantum Neural Network for Classification
- Building a hybrid QNN
- Training and evaluating a binary or multi-class classifier
- Tips for stabilizing training in QML
Additional materials:
- 4 Jupyter notebooks (from basic circuits to full QML examples)
- Setup scripts and environment guide
- AIFactory submission format guide
- Beginner-friendly glossary (qubit, gate, measurement, etc.)

8. Task, Data & Constraints

Task: Build a quantum model that classifies data points into given classes.
Data:
- A training + validation set is provided in the starter kit.
- A separate hidden test set is stored only on the organizer’s side and is never released to participants.
Constraints (details in the official problem document):
- Maximum number of qubits
- Maximum circuit depth / number of layers
- Allowed measurement scheme
- Any limits on number of parameters or evaluation time
Submissions must follow these constraints to be considered valid.

9. Submission Format (JSON with Quantum Circuit)

In this second edition, participants do not submit prediction CSV files. Instead, each submission is a JSON file describing:

{
  "measurements": [m0, m1],   // integer indices or configuration parameters
  "qasm": "OPENQASM 3.0; ..." // quantum circuit in QASM format
}

qasm

A string containing the quantum circuit in the agreed QASM format
Must compile and run in the official grading environment

measurements

A pair of integers (or a small list) specifying how to interpret measurement results
For example: which qubits to read out, or how to map bitstrings to class labels
The exact meaning will be documented in the problem statement.

Each upload should contain one JSON file per submission following the above structure. Invalid JSON or circuits that fail to run will receive a score of 0.

10. Automatic GPU-Based Grading Pipeline

All submissions are evaluated with an automatic grading system running on a GPU server. The pipeline (corresponding to the diagram you provided) works as follows:

Participant Upload
- The team uploads a JSON file of the form
  {"measurements": (int, int), "qasm": "<QASM string>"} via the AIFactory submission page.
Forward to Grading Service
- The JSON is sent to the Grade Submission function running in a GPU environment.
Hidden Test Inference
- The grading function:
  - Loads the hidden test dataset stored internally.
  - Builds the quantum circuit from the qasm string.
  - Applies the circuit to the entire hidden test set (using the defined data-encoding and measurement scheme).
  - Produces predicted labels for all test samples.
Metric Computation & Visualization
- Using the ground-truth labels, the system computes:
  - Test accuracy (acc) as the main ranking metric.
  - Phase diagrams for each class or decision region.
  - Confusion matrix and error summary (per-class accuracy, number of errors, etc.).
Leaderboard Update
- The accuracy score is automatically written to the database and immediately reflected on the public leaderboard.
Feedback to Participant
- On the submission result page, each team can see:
- Phase diagram plots showing decision boundaries / sample distribution.
- Score detail panel, including:
  - Overall accuracy
  - Per-class stats
  - Confusion matrix
  - Any warnings (invalid gates, constraint violations, etc.)
This pipeline ensures:
- No leakage of the hidden test set
- Reproducible, fair evaluation for all teams
- Rich visual feedback to help participants debug and improve their circuits.

11. Evaluation & Ranking

Preliminary Round

Primary metric: Classification accuracy on the hidden test set.
Submissions are ranked by accuracy in descending order.
Additional checks:
- Constraint violations (qubit count, depth, runtime) may lead to penalties or disqualification.
- Ties may be broken using secondary criteria such as:
  - Smaller circuit depth
  - Fewer parameters
  - Earlier submission time
    (Exact tie-breaking rules will be published in the detailed problem statement.)
The top teams on the preliminary online leaderboard advance to the final round (e.g., top 10 teams; the exact number will be announced).

Final Round (on Real Quantum Hardware)

Finalists will implement their best models under additional hardware-compatible constraints.
Circuits will be executed on a real quantum device provided by the sponsor.
Final ranking will be based on:
- Accuracy on hardware runs (averaged across shots)
- Robustness against noise
- Quality of explanation and presentation (if applicable; e.g., short talk or poster)

12. Prizes (Example, to be confirmed)

Planned awards (KRW):

🏆 Grand Prize (NORMA Award) – (TBD)
🥇 Excellent Award (AIFactory Award) – (TBD)
🥈 Outstanding Award A (Rigetti Computing Award) – (TBD)
🥈 Outstanding Award B (Jeonju University Award) – (TBD)
🏅 Encouragement Award (ATTO Research Award) – (TBD)

The final prizes and conditions will be confirmed on the challenge page.

13. Code of Conduct & IP

Participants must submit original work. Plagiarism or unauthorized copying of other teams’ solutions is strictly prohibited.
Sharing or attempting to reconstruct the hidden test data is not allowed.
By joining, participants agree that anonymized results and circuits may be used for:
- Research publications
- Educational materials
- Future tutorial examples (Specific IP and licensing terms will be clarified in the official rules.)

14. Contact

General inquiries & technical questions:
- AIFactory Q&A board on the challenge page
- Email: cs@aifactory.page
Sponsorship & partnership inquiries:
- Email: contact@aifactory.page