---
title: mathjson-alpha
canonical_url: "https://www.modelscope.cn/datasets/jondurbin/mathjson-alpha"
md_url: "https://www.modelscope.cn/datasets/jondurbin/mathjson-alpha.md"
repository: jondurbin/mathjson-alpha
last_updated: 2025-08-29
license: apache-2.0
storage_size: "25 MB"
downloads: 168
stars: 0
---

# mathjson-alpha

> mathjson-alpha - jondurbin 在 ModelScope 开源的数据集。This is a first pass at generating MathJSON formulations of math problems to allow deterministic calculations (via cortex-js/compute-engine).

jondurbin/mathjson-alpha 是 ModelScope 魔搭社区上的数据集，存储大小 25 MB，采用 apache-2.0 许可。

- **Repository**: jondurbin/mathjson-alpha
- **License**: apache-2.0
- **Storage size**: 25 MB
- **Downloads**: 168
- **Stars**: 0
- **Last updated**: 2025-08-29

Source: https://www.modelscope.cn/datasets/jondurbin/mathjson-alpha

---

This is a first pass at generating MathJSON formulations of math problems to allow deterministic calculations (via cortex-js/compute-engine).

LLMs are decent at problem formulation, but terrible at calculations, especially things like calculating cosine of R radians, floating point with high precision multiplication, etc.  Let's let LLMs do what they are good at and run the computation outside.