The reranker and generator are two critical components in the Retrieval-Augmented Generation (i.e., RAG) pipeline, responsible for ranking relevant documents and generating responses. However, due to differences in pre-training data and objectives, there is an inevitable gap between the documents ranked as relevant by the reranker and those required by the generator to support answering the query. To address this gap, we propose RADIO, a novel and practical preference alignment framework with RAtionale DIstillatiOn. Specifically, we first propose a rationale extraction method that leverages the reasoning capabilities of Large Language Models (LLMs) to extract the rationales necessary for answering the query. Subsequently, a rationale-based alignment process is designed to rerank the documents based on the extracted rationales, and fine-tune the reranker to align the preferences. We conduct extensive experiments on two tasks across three datasets to demonstrate the effectiveness of our approach compared to baseline methods. Our code is released online to ease reproduction.
@article{arxiv.2412.08519,
title = {Bridging Relevance and Reasoning: Rationale Distillation in Retrieval-Augmented Generation},
author = {Pengyue Jia and Derong Xu and Xiaopeng Li and Zhaocheng Du and Xiangyang Li and Yichao Wang and Yuhao Wang and Qidong Liu and Maolin Wang and Huifeng Guo and Ruiming Tang and Xiangyu Zhao},
journal= {arXiv preprint arXiv:2412.08519},
year = {2025}
}