Summary of Investigating Instruction Tuning Large Language Models on Graphs, by Kerui Zhu et al.
Investigating Instruction Tuning Large Language Models on Graphs
by Kerui Zhu, Bo-Wei Huang, Bowen Jin, Yizhu Jiao, Ming Zhong, Kevin Chang, Shou-De Lin, Jiawei Han
First submitted to arxiv on: 10 Aug 2024
Categories
- Main: Computation and Language (cs.CL)
- 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 paper explores the potential of Large Language Models (LLMs) in engaging with real-world graphs, aiming to understand how LLMs can effectively interact with graphs and generalize across various tasks. To investigate this, a dataset is constructed comprising 79 graph-related tasks from academic and e-commerce domains, featuring over 60,000 training and test samples. The study identifies the optimal graph representation that allows LLMs to comprehend complex graph structures, finding that JSON format consistently outperforms natural language and code formats across various LLMs and graph types. Additionally, key factors influencing the generalization abilities of instruction-tuned LLMs are evaluated by assessing their performance on in-domain and out-of-domain graph tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study looks at how Large Language Models (LLMs) can work with real-world graphs. Graphs are like maps that show connections between things. The researchers want to know if LLMs can learn to understand these maps and do different tasks on them. To test this, they created a big dataset with many graph-related problems from things like schoolwork and online shopping. They found that using JSON format to represent graphs helps LLMs understand complex map structures better than other ways of representing graphs. The researchers also looked at what makes it easier for LLMs to do different tasks on the maps, even if they haven’t seen those specific tasks before. |
Keywords
» Artificial intelligence » Generalization
