Summary of Distilling Large Language Models For Text-attributed Graph Learning, by Bo Pan et al.
Distilling Large Language Models for Text-Attributed Graph Learning
by Bo Pan, Zheng Zhang, Yifei Zhang, Yuntong Hu, Liang Zhao
First submitted to arxiv on: 19 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- 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 proposes a novel approach to learning Text-Attributed Graphs (TAGs) by synergizing Large Language Models (LLMs) and graph models. LLMs are powerful in few-shot and zero-shot TAG learning, but they suffer from scalability, cost, and privacy issues. To address these gaps, the authors propose distilling the power of LLMs into a local graph model for TAG learning. The framework involves letting LLMs teach an interpreter with rich textual rationale and then having a student model mimic the interpreter’s reasoning without relying on LLMs’ textual rationale. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about finding a new way to learn graphs that connect texts together, called Text-Attributed Graphs (TAGs). Right now, we rely too much on human helpers to label these graphs, which can be hard to get. Some super smart language models have shown they can learn TAGs all by themselves, but they’re limited because they need a lot of training data and energy. So, the authors came up with an idea: use these language models to teach another model how to think like them, without needing as much help or resources. |
Keywords
* Artificial intelligence * Few shot * Student model * Zero shot
