Ask Optimal Questions: Aligning Large Language Models with Retriever's Preference in Conversation

TL;DR

This paper introduces RetPO, a framework that optimizes language models to generate query rewrites aligned with retriever preferences, significantly improving conversational search performance.

Contribution

The paper proposes a novel retriever preference optimization framework and a large-scale RF dataset to enhance query rewriting in conversational search.

Findings

RetPO outperforms previous rewrite-then-retrieve methods.

Constructed RF dataset with over 410K query rewrites.

Fine-tuned LM shows superior results on benchmark tests.

Abstract

Conversational search, unlike single-turn retrieval tasks, requires understanding the current question within a dialogue context. The common approach of rewrite-then-retrieve aims to decontextualize questions to be self-sufficient for off-the-shelf retrievers, but most existing methods produce sub-optimal query rewrites due to the limited ability to incorporate signals from the retrieval results. To overcome this limitation, we present a novel framework RetPO (Retriever's Preference Optimization), which is designed to optimize a language model (LM) for reformulating search queries in line with the preferences of the target retrieval systems. The process begins by prompting a large LM to produce various potential rewrites and then collects retrieval performance for these rewrites as the retrievers' preferences. Through the process, we construct a large-scale dataset called RF collection, containing Retrievers' Feedback on over 410K query rewrites across 12K conversations. Furthermore, we fine-tune a smaller LM on this dataset to align it with the retrievers' feedback. Our resulting model demonstrates superiority on two benchmarks, surpassing the previous state-of-the-art performance of rewrite-then-retrieve approaches.