Summary of What Do Llms Need to Understand Graphs: a Survey Of Parametric Representation Of Graphs, by Dongqi Fu et al.
What Do LLMs Need to Understand Graphs: A Survey of Parametric Representation of Graphs
by Dongqi Fu, Liri Fang, Zihao Li, Hanghang Tong, Vetle I. Torvik, Jingrui He
First submitted to arxiv on: 16 Oct 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Machine Learning (cs.LG); Social and Information Networks (cs.SI)
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 survey investigates how large language models (LLMs) can be made to understand graph-based relational data. Graphs have been widely used in various application scenarios, such as molecule design and recommender systems. LLMs are reorganizing the AI community due to their expected reasoning and inference abilities. The authors identify two potential applications: distilling external knowledge bases to eliminate hallucination and breaking the context window limit for LLMs’ inference during the retrieval-augmentation generation process; and directly solving graph-based research tasks like protein design and drug discovery using graph data as input. However, feeding entire graph data to LLMs is impractical due to its complex topological structure, data size, and lack of effective and efficient semantic graph representations. The authors propose a parametric representation of graphs called “graph laws” that can be described by natural language for LLM’s understanding. Graph laws pre-define a set of parameters (e.g., degree, time, diameter) and identify their relationships and values by observing the topological distribution of real-world graph data. This representation can serve as the raw input for LLMs. The authors review previous studies on graph laws from multiple perspectives: macroscope and microscope of graphs, low-order and high-order graphs, static and dynamic graphs, different observation spaces, and newly proposed graph parameters. They also explore various real-world applications benefiting from the guidance of graph laws. Finally, they conclude with current challenges and future research directions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper explores how to make large language models (LLMs) understand graph-based relational data. Graphs are used in many areas like molecule design and recommender systems. LLMs can reason and infer well. The authors think that making LLMs understand graphs can be helpful for tasks like protein design and drug discovery. The problem is that feeding entire graph data to LLMs is hard because of the complexity of the data. The authors propose a way to represent graphs using “graph laws” which can be understood by LLMs. Graph laws are based on statistics and identify patterns in real-world graph data. The paper reviews previous studies on graph laws and shows how they can be used in many areas. It also talks about challenges and future directions for research. |
Keywords
» Artificial intelligence » Context window » Hallucination » Inference
