Summary of Enhancing Data Quality in Federated Fine-tuning Of Foundation Models, by Wanru Zhao et al.
Enhancing Data Quality in Federated Fine-Tuning of Foundation Models
by Wanru Zhao, Yaxin Du, Nicholas Donald Lane, Siheng Chen, Yanfeng Wang
First submitted to arxiv on: 7 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
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 Medium Difficulty summary: In the context of foundation model training, the reliance on public domain data is nearing exhaustion. To scale up, it’s essential to incorporate private domain data sources, which poses challenges in data quality control due to local training without sharing data. This paper proposes a data quality control pipeline for federated fine-tuning of foundation models, computing scores and determining a global threshold for improved performance. Our experiments show that this approach facilitates better model training, leading to enhanced reliability and effectiveness. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty summary: Right now, we’re running out of public domain data to train our machine learning models. To make progress, we need to use private data from different sources. The problem is that we can’t share this data because it’s not ours to share. This paper suggests a way to control the quality of this private data so we can still get good results without sharing it. Our tests show that this approach helps us train better models and makes them more reliable. |
Keywords
* Artificial intelligence * Fine tuning * Machine learning
