Summary of Sub-sa: Strengthen In-context Learning Via Submodular Selective Annotation, by Jian Qian et al.
Sub-SA: Strengthen In-context Learning via Submodular Selective Annotation
by Jian Qian, Miao Sun, Sifan Zhou, Ziyu Zhao, Ruizhi Hun, Patrick Chiang
First submitted to arxiv on: 8 Jul 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 The proposed Sub-SA method aims to reduce annotation costs while improving the quality of in-context examples for Large Language Models (LLMs). This is achieved by designing a submodular function that facilitates effective subset selection for annotation. The RPR regularization technique balances diversity and representativeness, allowing for efficient greedy search algorithm implementation. By leveraging this approach, the paper demonstrates improvements in ICL performance while minimizing time consumption. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large Language Models (LLMs) are great at predicting things when given examples to work with. But getting those example prompts can be a big job! To make it easier, researchers developed Sub-SA, a way to pick the best examples from a large pool of labeled ones. This makes the process faster and cheaper while still getting good results. The method uses math to help find the right examples and balances how diverse they are with how representative they are. This means we can use simple computer algorithms to make the selection. Overall, this new approach improves the performance of LLMs in learning from examples. |
Keywords
* Artificial intelligence * Regularization
