Summary of Unlocking Large Language Model’s Planning Capabilities with Maximum Diversity Fine-tuning, by Wenjun Li et al.
Unlocking Large Language Model’s Planning Capabilities with Maximum Diversity Fine-tuning
by Wenjun Li, Changyu Chen, Pradeep Varakantham
First submitted to arxiv on: 15 Jun 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
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 capabilities of large language models (LLMs) in planning tasks, which they often struggle with despite their impressive task-solving abilities. The authors investigate how fine-tuning affects LLMs’ performance and find that substantial fine-tuning can lead to good results, but at a high economic and computational cost. To address this challenge, the researchers propose the Maximum Diversity Fine-Tuning (MDFT) strategy, which improves sample efficiency in planning domains by encoding task instances with graph representations and selecting diverse samples. The proposed algorithm, MDFT-g, consistently outperforms existing baselines across multiple benchmark domains. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models can do many things well, but they often struggle to come up with good plans. This is a problem because planning is an important part of being able to solve many tasks. To see if we could help LLMs get better at planning, the authors fine-tuned them on thousands of specific examples and found that this helped. However, making these models better was expensive and used a lot of computing power. So, the researchers came up with a new way to fine-tune LLMs called Maximum Diversity Fine-Tuning (MDFT). This method helps the model learn from a smaller group of examples by choosing ones that are very different from each other. |
Keywords
» Artificial intelligence » Fine tuning
