Summary of On the Origin Of Llamas: Model Tree Heritage Recovery, by Eliahu Horwitz et al.
On the Origin of Llamas: Model Tree Heritage Recovery
by Eliahu Horwitz, Asaf Shul, Yedid Hoshen
First submitted to arxiv on: 28 May 2024
Categories
- Main: Machine Learning (cs.LG)
- 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 The paper proposes an innovative task called Model Tree Heritage Recovery (MoTHer Recovery), which aims to discover the origin of neural network models based on their weights. The authors define the Model Tree, a tree-like structure that represents the relationships between models, inspired by Darwin’s tree of life. They hypothesize that model weights encode this information and develop a method to decode the underlying tree structure given the weights. This task has applications in model authorship attribution and long-term goals akin to indexing the internet. The authors demonstrate their approach on various models, including Llama 2 and Stable Diffusion, successfully reconstructing complex Model Trees. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about how we can find out where neural network models come from. Right now, people share model weights online, but it’s hard to figure out which model came first or who trained another model. The authors want to solve this problem by looking at the model weights and finding patterns that tell us which models are related. They call this “Model Tree Heritage Recovery” and think it could be used to find out who created certain models, or even index all the different models online like we do with websites on the internet. |
Keywords
» Artificial intelligence » Diffusion » Llama » Neural network
