Summary of Familiarity-aware Evidence Compression For Retrieval-augmented Generation, by Dongwon Jung et al.
Familiarity-Aware Evidence Compression for Retrieval-Augmented Generation
by Dongwon Jung, Qin Liu, Tenghao Huang, Ben Zhou, Muhao Chen
First submitted to arxiv on: 19 Sep 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG)
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 Retrieval-augmented generation (RAG) is a technique that improves large language models (LMs) by incorporating external information from various sources. However, this approach often struggles with inconsistent and irrelevant data that can distract the LM from its tasks. To address this issue, researchers have proposed compressing the retrieved evidence to make it more familiar to the target model. A new compression technique called FaviComp (Familiarity-Aware Evidence Compression) aims to improve this process by making the compressed evidence more accessible to the target model while integrating both parametric and non-parametric knowledge. Experimental results show that FaviComp outperforms recent baselines in open-domain question-answering tasks, achieving accuracy improvements of up to 28.1% with high compression rates. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine trying to find the right information from many different sources to help a language model do its job better. This is called retrieval-augmented generation (RAG). However, it can be tricky when some of that information isn’t very helpful and gets in the way. To fix this problem, researchers have come up with a new way to compress the information so it’s easier for the language model to understand. They call it FaviComp. It helps the model use both the helpful and unhelpful information better. In tests, FaviComp did much better than other methods, getting answers right more often while still using most of the information. |
Keywords
» Artificial intelligence » Language model » Question answering » Rag » Retrieval augmented generation
