PRCA: Fitting Black-Box Large Language Models for Retrieval Question Answering via Pluggable Reward-Driven Contextual Adapter

TL;DR

This paper introduces PRCA, a pluggable adapter that improves retrieval-based question answering with large black-box language models by using reinforcement learning to refine retrieved information.

Contribution

The paper presents a novel pluggable adapter, PRCA, that enhances black-box LLMs in retrieval question answering without fine-tuning the models.

Findings

PRCA improves ReQA performance by up to 20%.

PRCA effectively refines retrieved information using reinforcement learning.

PRCA enables fitting black-box LLMs into existing frameworks.

Abstract

The Retrieval Question Answering (ReQA) task employs the retrieval-augmented framework, composed of a retriever and generator. The generator formulates the answer based on the documents retrieved by the retriever. Incorporating Large Language Models (LLMs) as generators is beneficial due to their advanced QA capabilities, but they are typically too large to be fine-tuned with budget constraints while some of them are only accessible via APIs. To tackle this issue and further improve ReQA performance, we propose a trainable Pluggable Reward-Driven Contextual Adapter (PRCA), keeping the generator as a black box. Positioned between the retriever and generator in a Pluggable manner, PRCA refines the retrieved information by operating in a token-autoregressive strategy via maximizing rewards of the reinforcement learning phase. Our experiments validate PRCA's effectiveness in enhancing ReQA performance on three datasets by up to 20% improvement to fit black-box LLMs into existing frameworks, demonstrating its considerable potential in the LLMs era.