Summary of Openmathinstruct-1: a 1.8 Million Math Instruction Tuning Dataset, by Shubham Toshniwal et al.
OpenMathInstruct-1: A 1.8 Million Math Instruction Tuning Dataset
by Shubham Toshniwal, Ivan Moshkov, Sean Narenthiran, Daria Gitman, Fei Jia, Igor Gitman
First submitted to arxiv on: 15 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed paper constructs an open-source math instruction tuning dataset using a permissively licensed language model, addressing the limitation of relying on closed-source models with restrictive licenses. The dataset, OpenMathInstruct-1, contains 1.8M problem-solution pairs synthesized using the Mixtral model and popular math reasoning benchmarks GSM8K and MATH. The paper’s best model, OpenMath-CodeLlama-70B, achieves competitive scores on these benchmarks, outperforming gpt-distilled models. This work aims to promote the use of open-source language models in large-scale math instruction tuning datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper creates a big dataset for teaching math using a special computer model that can be used freely by anyone. They made this dataset using two popular math tests and a helpful language model called Mixtral. The new model they trained is really good at solving math problems, almost as good as other models that are not open-source. This means we can use these free models to help teach math in the future. |
Keywords
* Artificial intelligence * Gpt * Instruction tuning * Language model
