Summary of Do Influence Functions Work on Large Language Models?, by Zhe Li et al.
Do Influence Functions Work on Large Language Models?
by Zhe Li, Wei Zhao, Yige Li, Jun Sun
First submitted to arxiv on: 30 Sep 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 A systematic study evaluates influence functions on large language models (LLMs), finding poor performance across multiple tasks due to approximation errors, uncertain convergence during fine-tuning, and limitations in defining influence functions for LLMs. The research highlights the challenges of applying traditional machine learning techniques to LLMs, suggesting a need for alternative approaches to identify influential samples. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Influence functions help us understand how individual data points affect model predictions. Researchers looked at these functions on large language models (LLMs) and found they don’t work well in most cases. This is because estimating the influence function for LLMs is very hard, fine-tuning can be unpredictable, and the way we define influence functions doesn’t fit with how LLMs behave. The study shows that traditional machine learning methods might not be suitable for large language models, so new ways are needed to figure out which data points matter. |
Keywords
» Artificial intelligence » Fine tuning » Machine learning
