Summary of How to Train Data-efficient Llms, by Noveen Sachdeva et al.
How to Train Data-Efficient LLMs
by Noveen Sachdeva, Benjamin Coleman, Wang-Cheng Kang, Jianmo Ni, Lichan Hong, Ed H. Chi, James Caverlee, Julian McAuley, Derek Zhiyuan Cheng
First submitted to arxiv on: 15 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 A novel approach to pre-training large language models (LLMs) is proposed in this paper, focusing on optimizing the Pareto frontier of model quality and training resource/data consumption. The study explores data-efficient techniques, including Ask-LLM, which leverages instruction-tuned LLMs’ zero-shot reasoning capabilities to assess training example quality, and Density sampling, a coverage-focused method that models data distribution. A comparison of 19 samplers reveals Ask-LLM and Density as the top performers in their respective categories. The results demonstrate that models trained on Ask-LLM data outperform full-data training, even when rejecting 90% of the original dataset, while converging up to 70% faster. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models (LLMs) are expensive to train. This paper helps make them more affordable by finding ways to use less data and training resources. The researchers developed two new methods: Ask-LLM and Density sampling. Ask-LLM uses special language models that can reason without needing examples, while Density sampling tries to select diverse samples from the available data. They tested 19 different methods and found that Ask-LLM and Density were the best at improving model performance with less training data. |
Keywords
* Artificial intelligence * Zero shot
