P1: Mastering Physics Olympiads with Reinforcement Learning

2025-11-17

AI Summary

Existing large language models (LLMs) exhibit limited scientific reasoning capabilities on Olympiad-level physics problems, particularly in complex physical modeling, multi-step deductive reasoning, and systematic application of conservation laws. Method: We propose a pure reinforcement learning (RL) training paradigm and introduce PhysicsMinions—a lightweight, multi-agent framework enabling role-based collaborative reasoning and physics-knowledge-guided stepwise problem solving. Contribution/Results: Our released P1 series models (e.g., P1-235B-A22B) achieve the first open-source LLM gold medal at IPhO 2025. Across 13 international physics competitions, they attain 12 gold and 1 silver medals—ranking first in both aggregate score and average performance, significantly surpassing state-of-the-art baselines. This work establishes a reproducible, scalable paradigm for scientific AI in physics reasoning.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Recent progress in large language models (LLMs) has moved the frontier from puzzle-solving to science-grade reasoning-the kind needed to tackle problems whose answers must stand against nature, not merely fit a rubric. Physics is the sharpest test of this shift, which binds symbols to reality in a fundamental way, serving as the cornerstone of most modern technologies. In this work, we manage to advance physics research by developing large language models with exceptional physics reasoning capabilities, especially excel at solving Olympiad-level physics problems. We introduce P1, a family of open-source physics reasoning models trained entirely through reinforcement learning (RL). Among them, P1-235B-A22B is the first open-source model with Gold-medal performance at the latest International Physics Olympiad (IPhO 2025), and wins 12 gold medals out of 13 international/regional physics competitions in 2024/2025. P1-30B-A3B also surpasses almost all other open-source models on IPhO 2025, getting a silver medal. Further equipped with an agentic framework PhysicsMinions, P1-235B-A22B+PhysicsMinions achieves overall No.1 on IPhO 2025, and obtains the highest average score over the 13 physics competitions. Besides physics, P1 models also present great performance on other reasoning tasks like math and coding, showing the great generalibility of P1 series.

Problem

• Developing large language models for physics reasoning
• Solving Olympiad-level physics problems using reinforcement learning
• Advancing science-grade reasoning beyond puzzle-solving capabilities

Innovation

• Training physics reasoning models with reinforcement learning
• Achieving gold-medal performance in Physics Olympiads
• Integrating agentic frameworks for enhanced problem-solving capabilities