Summary of In2core: Leveraging Influence Functions For Coreset Selection in Instruction Finetuning Of Large Language Models, by Ayrton San Joaquin et al.
In2Core: Leveraging Influence Functions for Coreset Selection in Instruction Finetuning of Large Language Models
by Ayrton San Joaquin, Bin Wang, Zhengyuan Liu, Nicholas Asher, Brian Lim, Philippe Muller, Nancy F. Chen
First submitted to arxiv on: 7 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Machine Learning (stat.ML)
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 the In2Core algorithm, which selects a coreset for fine-tuning Large Language Models (LLMs) by analyzing the correlation between training and evaluation samples with a trained model. The approach uses internal gradients to estimate this relationship, ranking the contribution of each training point. To enhance efficiency, the authors propose an optimization to compute influence functions with reduced layers, achieving similar accuracy. The algorithm is applied to instruction fine-tuning data, demonstrating comparable performance using only 50% of the original training data. Additionally, the paper explores the use of influence functions to analyze model coverage for test points. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research aims to make it easier and more affordable to improve Large Language Models by reducing the amount of data needed. The team developed a new way to select the most important information from large datasets, which can help train models using less data while still achieving good results. By doing this, they hope to make it possible for more people to contribute to open-source language model projects, even if they don’t have access to powerful computers. |
Keywords
» Artificial intelligence » Fine tuning » Language model » Optimization
