Improving Retrieval Augmented Language Model with Self-Reasoning

TL;DR

This paper introduces a self-reasoning framework for Retrieval-Augmented Language Models that enhances their reliability and traceability by leveraging internal reasoning trajectories, leading to improved performance on knowledge-intensive tasks.

Contribution

The paper proposes a novel self-reasoning framework that improves RALMs' reliability and traceability by using reasoning trajectories generated by the LLM itself, addressing issues of irrelevant retrieval and lack of citations.

Findings

Outperforms existing state-of-the-art models on multiple datasets

Achieves comparable performance to GPT-4 with only 2,000 training samples

Enhances model reliability and traceability through self-reasoning trajectories

Abstract

The Retrieval-Augmented Language Model (RALM) has shown remarkable performance on knowledge-intensive tasks by incorporating external knowledge during inference, which mitigates the factual hallucinations inherited in large language models (LLMs). Despite these advancements, challenges persist in the implementation of RALMs, particularly concerning their reliability and traceability. To be specific, the irrelevant document retrieval may result in unhelpful response generation or even deteriorate the performance of LLMs, while the lack of proper citations in generated outputs complicates efforts to verify the trustworthiness of the models. To this end, we propose a novel self-reasoning framework aimed at improving the reliability and traceability of RALMs, whose core idea is to leverage reasoning trajectories generated by the LLM itself. The framework involves constructing self-reason trajectories with three processes: a relevance-aware process, an evidence-aware selective process, and a trajectory analysis process. We have evaluated our framework across four public datasets (two short-form QA datasets, one long-form QA dataset, and one fact verification dataset) to demonstrate the superiority of our method, which can outperform existing state-of-the-art models and can achieve comparable performance with GPT-4, while only using 2,000 training samples.