PhySense: Principle-Based Physics Reasoning Benchmarking for Large Language Models

TL;DR

PhySense is a new benchmark designed to evaluate large language models' ability to perform principle-based physics reasoning, revealing their struggles to emulate expert-like, concise, and interpretable problem-solving approaches.

Contribution

The paper introduces PhySense, a physics reasoning benchmark that highlights the gap between LLMs and human experts in applying core principles for scientific reasoning.

Findings

LLMs often fail to follow principle-based reasoning paths.

Current LLMs generate lengthy, opaque solutions instead of concise, principle-driven ones.

PhySense exposes the limitations of LLMs in physics reasoning.

Abstract

Large language models (LLMs) have rapidly advanced and are increasingly capable of tackling complex scientific problems, including those in physics. Despite this progress, current LLMs often fail to emulate the concise, principle-based reasoning characteristic of human experts, instead generating lengthy and opaque solutions. This discrepancy highlights a crucial gap in their ability to apply core physical principles for efficient and interpretable problem solving. To systematically investigate this limitation, we introduce PhySense, a novel principle-based physics reasoning benchmark designed to be easily solvable by experts using guiding principles, yet deceptively difficult for LLMs without principle-first reasoning. Our evaluation across multiple state-of-the-art LLMs and prompt types reveals a consistent failure to align with expert-like reasoning paths, providing insights for developing AI systems with efficient, robust and interpretable principle-based scientific reasoning.