LAR: LLM Assisted Retrieval
Abstract: Large language models (LLMs), have demonstrated significant success in natural language understanding and generation tasks. In this work, we propose LAR (Large language model Assisted Retrieval) to harness LLMs towards enhancing the effectiveness of retrieval models, thereby improving the relevance of information retrieval from datasets.
Our approach augments a retriever engine by incorporating a subsequent refinement step to the query, utilizing an LLM. This approach showcases the potential of combining retrieval models with LLMs to advance information retrieval systems. We demonstrate the efficacy of LAR through extensive evaluations, specifically showing enhanced performance on the BEIR retrieval benchmark. Furthermore, our methodology exhibits notable improvements on downstream tasks such as question answering, as demonstrated on the NarrativeQA dataset.
Paper Type: Short
Research Area: Information Retrieval and Text Mining
Research Area Keywords: passage retrieval,dense retrieval,re-ranking,zero/few-shot extraction,retrieval-augmented models
Contribution Types: Publicly available software and/or pre-trained models
Languages Studied: English
Submission Number: 5288
Loading