Summary of Rare: Retrieval Augmented Retrieval with In-context Examples, by Atula Tejaswi et al.
RARe: Retrieval Augmented Retrieval with In-Context Examples
by Atula Tejaswi, Yoonsang Lee, Sujay Sanghavi, Eunsol Choi
First submitted to arxiv on: 26 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
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 This paper explores whether incorporating in-context examples, a technique commonly used in decoder-only language models, can enhance the performance of embedding models in retrieval tasks. Unlike in LLMs, simply prepending query-document pairs to the target query during inference time does not yield expected results. The authors introduce RARe, a straightforward approach that finetunes a pre-trained model with semantically similar queries to the target query. This method can be applied to various base architectures and consistently achieves performance gains of up to +2.72% nDCG across open-domain retrieval datasets like BeIR and RAR-b. The study finds RARe exhibits stronger out-of-domain generalization compared to models using queries without in-context examples, similar to what is seen for in-context learning in LLMs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how adding extra information to the query can help language models find relevant answers. They found that just adding this extra information isn’t enough and created a new way to use it called RARe. This method helps the model learn from examples that are similar to what it’s trying to find. It works with different types of models and makes them better at finding answers. The results show that RARe does even better than other methods when used in new situations. |
Keywords
» Artificial intelligence » Decoder » Domain generalization » Embedding » Inference
