SymPyBench: A Dynamic Benchmark for Scientific Reasoning with Executable Python Code

TL;DR

SymPyBench is a comprehensive, executable Python-based benchmark with diverse physics problems designed to evaluate and improve scientific reasoning in AI models, introducing new metrics for variability and uncertainty.

Contribution

It presents a large-scale, parameterized physics problem benchmark with novel evaluation metrics, enabling dynamic testing of reasoning skills in AI models.

Findings

State-of-the-art models show strengths in some reasoning tasks.

The benchmark reveals limitations in models' consistency and uncertainty handling.

New metrics provide deeper insights into model reasoning variability.

Abstract

We introduce, a large-scale synthetic benchmark of 15,045 university-level physics problems (90/10% train/test split). Each problem is fully parameterized, supporting an effectively infinite range of input configurations, and is accompanied by structured, step-by-step reasoning and executable Python code that produces the ground-truth solution for any parameter set. The benchmark contains three question types: MC-Symbolic (multiple-choice with symbolic options), MC-Numerical (multiple-choice with numerical options), and free-form (open-ended responses). These diverse formats test complementary reasoning skills. By leveraging the dynamic, code-driven nature of the benchmark, we introduce three novel evaluation metrics in addition to standard accuracy: Consistency Score, Failure Rate, and Confusion Rate, that quantify variability and uncertainty across problem variants. Experiments with state-of-the-art instruction-tuned language models reveal both strengths and limitations in scientific reasoning, positioning SymPyBench as a foundation for developing more robust and interpretable reasoning systems