Improving Zero-shot LLM Re-Ranker with Risk Minimization
This work addresses a specific bias problem in zero-shot LLM re-ranking for RAG systems, offering an incremental improvement for information retrieval and QA applications.
The paper tackles the bias in using LLMs as unsupervised query likelihood models for document re-ranking in RAG systems by introducing the UR^3 framework, which uses Bayesian decision theory to mitigate estimation bias and improves Top-1 accuracy in QA tasks with fewer documents.
In the Retrieval-Augmented Generation (RAG) system, advanced Large Language Models (LLMs) have emerged as effective Query Likelihood Models (QLMs) in an unsupervised way, which re-rank documents based on the probability of generating the query given the content of a document. However, directly prompting LLMs to approximate QLMs inherently is biased, where the estimated distribution might diverge from the actual document-specific distribution. In this study, we introduce a novel framework, $\mathrm{UR^3}$, which leverages Bayesian decision theory to both quantify and mitigate this estimation bias. Specifically, $\mathrm{UR^3}$ reformulates the problem as maximizing the probability of document generation, thereby harmonizing the optimization of query and document generation probabilities under a unified risk minimization objective. Our empirical results indicate that $\mathrm{UR^3}$ significantly enhances re-ranking, particularly in improving the Top-1 accuracy. It benefits the QA tasks by achieving higher accuracy with fewer input documents.