IBM Qiskit Hackathon - Bangladesh

Dates

All times are local time at the venue.

Eligibility

• Open to [eligible participants, to be confirmed by the organizers].
• Teams of [minimum] to [maximum] members. Each participant may join only one team.
• Every team member must register on this Devpost page and be listed on the team's project.
• Teams may start working from home after the release at 10:00 AM and should join the in-person session at the venue.

Project and Submission Requirements

• Your own work. Submit your team's own work. You may discuss ideas with other teams, but do not share code.
• AI tools. You may use AI assistants to help you understand concepts and theory, for example how the CHSH inequality or the Jordan-Wigner mapping works. The code you submit must be written by your team; do not submit code produced by AI tools. Judges may ask a team to explain its code.
• Protected code. Do not edit, delete, duplicate or reorder cells marked # Provided Code: Do Not Modify, and do not change the names or parameter lists of the required functions. The package README.md lists the consequences of each kind of change.
• Fair play. Do not hardcode results, try to read hidden test data, or bypass the provided simulation helpers and resource budgets. Results must come from running your own implementation.
• Packages. Use only the Python standard library and the packages in requirements.txt. Grading runs without network access, so packages installed from inside a notebook are not available.
• Reproducibility. Every notebook must run from top to bottom in a fresh kernel within the time limits stated in the package README.md.
• Citations. Cite papers, tutorials and documentation that you use.

Prizes

[To be filled in by the organizers]

Judging Criteria and Winner Selection

Each challenge is worth 100 points; the maximum total is 300 points.

• Automatic evaluation (90 points per challenge). After the deadline, every submitted notebook is executed again from top to bottom in a clean, isolated environment with a fresh random seed. Saved outputs in your files are ignored. Your functions are then called with hidden inputs (other keys, angles, noise levels and molecular geometries) and compared with exact or statistical expectations. Statistical checks use generous tolerances, so a correct implementation passes with overwhelming probability, whatever the random seed.
• Scientific interpretation (10 points per challenge). Judges assess your written answers, experiments and figures with a structured review form that is identical for every team.
• Challenge 3 accuracy. Part of the VQE score depends on how close your energies come to the exact reference values at the hidden bond lengths, on a continuous scale, so better methods earn more points.

The complete point breakdown for every task is published in Section 11 of each notebook.

Winner selection. Teams are ranked by total points. Ties are broken, in order, by: (1) the higher Challenge 3 score; (2) the smaller mean error of the VQE energies at the hidden bond lengths; (3) fewer two-qubit gates in the VQE circuits; (4) fewer total shots used in the VQE runs; (5) the higher Challenge 2 score. Teams that are still tied share the rank. Rule violations are reviewed by the judges, who may set the affected challenge to zero points or disqualify a team for deliberate cheating. Teams can appeal through the organizers before the final ranking is published.