Summary of Infinitymath: a Scalable Instruction Tuning Dataset in Programmatic Mathematical Reasoning, by Bo-wen Zhang et al.
InfinityMATH: A Scalable Instruction Tuning Dataset in Programmatic Mathematical Reasoning
by Bo-Wen Zhang, Yan Yan, Lin Li, Guang Liu
First submitted to arxiv on: 9 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 recent advancements in Chain-of-Thoughts (CoT) and Program-of-Thoughts (PoT) methods have significantly enhanced the mathematical reasoning capabilities of language models. To facilitate their integration into instruction tuning datasets, researchers introduced InfinityMATH, a scalable dataset for programmatic mathematical reasoning. The construction pipeline decouples numbers from mathematical problems, synthesizing number-independent programs that enable efficient and flexible scaling. Fine-tuning experiments with open-source language and code models demonstrate the practical benefits of InfinityMATH. The fine-tuned models showed significant relative improvements on in-domain and out-of-domain benchmarks, ranging from 184.7% to 514.3%. Additionally, these models exhibited high robustness on the GSM8K+ and MATH+ benchmarks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary InfinityMATH is a new way to teach language models math. It’s like a big book of problems that helps them get better at solving math questions. Right now, it’s hard to make really big datasets for teaching math because it takes a lot of time and computers. InfinityMATH solves this problem by making a special kind of math problem that doesn’t rely on specific numbers. This makes it easy to scale up the dataset and use it with different models. When they tested InfinityMATH, the results were impressive! The language models got much better at solving math problems, and they even did well on new, harder tests. |
Keywords
» Artificial intelligence » Fine tuning » Instruction tuning
