Task refinement improves retrieval augmented language model answers
TRIO: Task Refinement with Individual Optimization for Retrieval-Augmented Generation
Computation and Language
Summary
Getting language models to provide accurate answers can be hard because they sometimes don't retrieve the right information or make things up. The paper introduces a method called TRIO that uses three agents working together to improve this process. One agent expands unclear questions to get better info, another judges if the retrieved information is good enough, and a third refines the final answer based on this feedback. In tests, this approach improved the quality of answers compared to previous methods.
What this means in practice
- •For chatbot developers: Improve chatbot accuracy by refining question understanding and retrieval before generating responses.
- •For knowledge base maintainers: Enhance retrieval systems to provide better expanded queries and judge information sufficiency for automated answer generation.
Authors
Yingtao Ren, Xiao Luo, Yiwei Fu, Jinzhao Zhou, Yu-Cheng Chang, Chin-Teng Lin
Abstract
This paper studies retrieval-augmented generation (RAG), which leverages external knowledge to improve the performance of language models. Despite the remarkable progress, current RAG approaches are still far from satisfactory due to inadequate retrieval and potential hallucinations. Towards this end, we propose a novel approach named Task Refinement with Individual Optimization (TRIO) for RAG. The core idea of our TRIO is to equip three well-designed agents with an iterative refinement paradigm by aligning their individual optimization objectives. Specifically, we first expand ambiguous queries with a query extension agent and quantify the reward based on the improved answer quality, which can be utilized to refine our agent for higher expansion efficiency. To enhance the retrieval quality, we include a unanimity voting to annotate the expanded query as insufficient or sufficient, and their generative outcomes are utilized as ground truth to optimize the judge agent. To further improve evidence utilization and answer accuracy, an answer agent is optimized with dynamic matching-based rewards with curriculum learning for final outputs. Extensive experiments across multiple benchmark datasets validate the effectiveness of the proposed TRIO in comparison with state-of-the-art approaches. Our code is available at https://github.com/yingtaoren/TRIO.