We present Gecko, a compact and versatile text embedding model. Gecko achieves strong retrieval performance by leveraging a key idea: distilling knowledge from large language models (LLMs) into a retriever. Our two-step distillation process begins with generating diverse, synthetic paired data using an LLM. Next, we further refine the data quality by retrieving a set of candidate passages for each query, and relabeling the positive and hard negative passages using the same LLM. The effectiveness of our approach is demonstrated by the compactness of the Gecko. On the Massive Text Embedding Benchmark (MTEB), Gecko with 256 embedding dimensions outperforms all existing entries with 768 embedding size. Gecko with 768 embedding dimensions achieves an average score of 66.31, competing with 7x larger models and 5x higher dimensional embeddings.
@article{arxiv.2403.20327,
title = {Gecko: Versatile Text Embeddings Distilled from Large Language Models},
author = {Jinhyuk Lee and Zhuyun Dai and Xiaoqi Ren and Blair Chen and Daniel Cer and Jeremy R. Cole and Kai Hui and Michael Boratko and Rajvi Kapadia and Wen Ding and Yi Luan and Sai Meher Karthik Duddu and Gustavo Hernandez Abrego and Weiqiang Shi and Nithi Gupta and Aditya Kusupati and Prateek Jain and Siddhartha Reddy Jonnalagadda and Ming-Wei Chang and Iftekhar Naim},
journal= {arXiv preprint arXiv:2403.20327},
year = {2024}
}