Date
AI Overview
Why it matters
Mastering International Physics Olympiad problems pushes automated reasoning forward. While prior AI models reached gold-medal levels, they trailed top human contestants, making advanced scientific reasoning a major frontier.
Study at a glance
- Question: Can an AI agent solve high-level International Physics Olympiad theory problems with perfect accuracy?
- Setting and data: Three theoretical problems from the 2025 International Physics Olympiad.
- Method: Model or benchmark evaluation using an agentic workflow with parallel thinking, agentic vision, and expert human grading.
- AI tools: Gemini 3.1 Pro Preview
Key findings
- The Gemini-powered agent achieved a perfect score of 30 out of 30 across five independent evaluation runs on the 2025 International Physics Olympiad theory exam.
- Parallel thinking synthesis corrected hidden sign and pre-factor calculation mistakes that occurred in raw single-solution attempts.
- A dedicated measurement component using Python-based agentic vision resolved errors stemming from native visual estimation on diagrams.
Summary drafted with AI from the full paper and may contain errors. See the full paper for details.
The International Physics Olympiad (IPhO) is the world's most prestigious and renowned physics competition for pre-university students. IPhO problems require complex reasoning based on deep understanding of physical principles in a standard general physics curriculum. On IPhO 2025 theory problems, while gold medal performance by AI models was reported previously, it falls behind the best human contestant. Here we build a simple agent with Gemini 3.1 Pro Preview. We run it five times and it achieved a perfect score every time. However, data contamination could occur because Gemini 3.1 Pro Preview was released after the competition.
Research Categories
Who age?
Why use AI?
Study design
AI tools
Publication Source
Publisher
arXiv
