When Do Program-of-Thoughts Work for Reasoning?

In the realm of embodied artificial intelligence, the reasoning capabilities of Large Language Models (LLMs) play a pivotal role. Although there are effective methods like program-of-thought prompting for LLMs which uses programming language to tackle complex reasoning tasks, the specific impact of...

Full description

Saved in:
Bibliographic Details
Published inarXiv.org
Main Authors Bi, Zhen, Zhang, Ningyu, Jiang, Yinuo, Deng, Shumin, Zheng, Guozhou, Chen, Huajun
Format Paper
LanguageEnglish
Published Ithaca Cornell University Library, arXiv.org 18.12.2023
Subjects
Algorithms
Artificial intelligence
Empirical analysis
Large language models
Programming languages
Reasoning
Task complexity
Online AccessGet full text

Cover

More Information
Summary:In the realm of embodied artificial intelligence, the reasoning capabilities of Large Language Models (LLMs) play a pivotal role. Although there are effective methods like program-of-thought prompting for LLMs which uses programming language to tackle complex reasoning tasks, the specific impact of code data on the improvement of reasoning capabilities remains under-explored. To address this gap, we propose complexity-impacted reasoning score (CIRS), which combines structural and logical attributes, to measure the correlation between code and reasoning abilities. Specifically, we use the abstract syntax tree to encode the structural information and calculate logical complexity by considering the difficulty and the cyclomatic complexity. Through an empirical analysis, we find not all code data of complexity can be learned or understood by LLMs. Optimal level of complexity is critical to the improvement of reasoning abilities by program-aided prompting. Then we design an auto-synthesizing and stratifying algorithm, and apply it to instruction generation for mathematical reasoning and code data filtering for code generation tasks. Extensive results demonstrates the effectiveness of our proposed approach. Code will be integrated into the EasyInstruct framework at https://github.com/zjunlp/EasyInstruct.
ISSN:2331-8422