Summary of On the Empirical Complexity Of Reasoning and Planning in Llms, by Liwei Kang et al.
On the Empirical Complexity of Reasoning and Planning in LLMs
by Liwei Kang, Zirui Zhao, David Hsu, Wee Sun Lee
First submitted to arxiv on: 17 Apr 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: 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 This paper explores the surprisingly good performance of Chain-of-thought (CoT) and Tree-of-thought (ToT) techniques in Large Language Models (LLMs) for complex reasoning tasks. The authors investigate why these methods work well by conducting experimental case studies and linking their benefits to established machine learning principles. They tested 6 reasoning tasks, ranging from grade school math to Blocksworld, and found that task decomposition is crucial for both CoT and ToT, breaking down complex problems into simpler steps with low sample complexity. Additionally, the study reveals that ToT outperforms CoT on computationally hard tasks due to its more sophisticated tree structure. These findings provide valuable guidelines for using LLMs in practical reasoning tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how Chain-of-thought (CoT) and Tree-of-thought (ToT) help Large Language Models (LLMs) solve complex problems. The authors want to know why these methods work so well, so they did some experiments and looked at established machine learning ideas. They tried 6 different types of reasoning tasks and found that breaking down big problems into smaller steps helps both CoT and ToT. They also discovered that the more complicated tree structure of ToT is better for hard problems than the simpler chain used in CoT. This research gives useful tips on how to use LLMs to solve real-world problems. |
Keywords
» Artificial intelligence » Machine learning
