Summary of Large Language Models For Constrained-based Causal Discovery, by Kai-hendrik Cohrs et al.
Large Language Models for Constrained-Based Causal Discovery
by Kai-Hendrik Cohrs, Gherardo Varando, Emiliano Diaz, Vasileios Sitokonstantinou, Gustau Camps-Valls
First submitted to arxiv on: 11 Jun 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Computation and Language (cs.CL)
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 potential of Large Language Models (LLMs) in constructing causal graphs, which are essential for understanding complex systems like economies, brains, and climates. The authors propose framing conditional independence queries as prompts to LLMs and using the PC algorithm with their answers. They demonstrate that the performance of this approach on systems with known causal graphs is variable, but improve it through a statistical-inspired voting schema. Furthermore, they show that knowledge-based CIT can eventually become a complementary tool for data-driven causal discovery. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how to create diagrams that show cause-and-effect relationships in complex things like economies and brains. Right now, we have two ways to do this: one uses lots of data, but might not work if we don’t have enough data or if the data isn’t good enough. The other way relies on experts who know a lot about the subject, but it takes them a long time to come up with an answer. This paper explores using big language models to help create these diagrams instead of relying solely on data or experts. It shows that this approach can be helpful, and even proposes ways to make it better. |
