Summary of Small Language Model As Data Prospector For Large Language Model, by Shiwen Ni et al.
Small Language Model as Data Prospector for Large Language Model
by Shiwen Ni, Haihong Wu, Di Yang, Qiang Qu, Hamid Alinejad-Rokny, Min Yang
First submitted to arxiv on: 13 Dec 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 Medium Difficulty summary: The quality of instruction data is crucial for fine-tuned Large Language Models (LLMs). Building upon previous work on NUGGETS, which identifies high-quality data by selecting individual examples that improve task performance after being learned as one-shot instances, this paper proposes SuperNUGGETS. This improved variant optimizes efficiency and performance using a small language model (SLM) to filter outstanding one-shot instances and refine predefined tests. The results show that SuperNUGGETS’ performance decreases by only 1-2% compared to NUGGETS, while increasing efficiency by a factor of 58. This improvement in utility value is due to significantly lower resource consumption. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty summary: This paper talks about how the quality of instructions affects big language models. It’s like giving a good teacher a great lesson plan – it makes a big difference! The authors took an idea from before called NUGGETS and made it better, called SuperNUGGETS. They wanted to make it more efficient and do its job just as well. And guess what? It worked! SuperNUGGETS only did a little bit worse than the original, but used way less resources. That’s like having a superpower for language models! |
Keywords
» Artificial intelligence » Language model » One shot
