InstUPR : Instruction-based Unsupervised Passage Reranking with Large Language Models

TL;DR

InstUPR is an unsupervised passage reranking method utilizing instruction-tuned large language models, employing soft score aggregation and pairwise reranking, which outperforms existing unsupervised and instruction-tuned baselines on the BEIR benchmark.

Contribution

This work introduces a novel unsupervised reranking approach that leverages instruction-following LLMs without additional fine-tuning, using a soft score aggregation and pairwise reranking technique.

Findings

Outperforms unsupervised baselines on BEIR benchmark

Surpasses instruction-tuned rerankers in effectiveness

Demonstrates the power of instruction-following LLMs for reranking

Abstract

This paper introduces InstUPR, an unsupervised passage reranking method based on large language models (LLMs). Different from existing approaches that rely on extensive training with query-document pairs or retrieval-specific instructions, our method leverages the instruction-following capabilities of instruction-tuned LLMs for passage reranking without any additional fine-tuning. To achieve this, we introduce a soft score aggregation technique and employ pairwise reranking for unsupervised passage reranking. Experiments on the BEIR benchmark demonstrate that InstUPR outperforms unsupervised baselines as well as an instruction-tuned reranker, highlighting its effectiveness and superiority. Source code to reproduce all experiments is open-sourced at https://github.com/MiuLab/InstUPR