International Theoretical Physics Olympiad

Under § 6 of the ITPO Statutes (Version 2.1, ratified January 2024), the Secretariat is authorized to establish a separate "Machine Intelligence Track" for the evaluation of artificial intelligence systems on ITPO competition problems.

Results from this track are ranked independently and do not affect the ranking or medal distribution of human participants.

AI systems are evaluated on the same theoretical physics problems administered to human participants. The problem archive (2016–2021) is publicly available and may be used for research purposes.

Problems require multi-step analytical derivations from first principles in areas including classical mechanics, quantum mechanics, statistical physics, electrodynamics, and relativity.

Submissions are evaluated by the same jury that assesses human solutions, applying identical grading rubrics. Evaluation is based on:

• Correctness of physical reasoning and final results
• Completeness of intermediate derivation steps
• Mathematical rigor and proper use of notation
• Clarity and logical structure of the solution

The ITPO problem archive has been used by research institutions to evaluate reasoning capabilities in large language models and symbolic AI systems.

Researchers may use the problem set for benchmarking purposes without prior authorization. Publication of results should cite the ITPO archive and specify the competition year and problem numbers used.

Organizations wishing to submit AI system solutions for official evaluation may contact the Secretariat. Submissions must include a technical specification of the system, computational resources used, and any human intervention in the solution process.

Official results are published in a separate leaderboard maintained by the Secretariat.