arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2402.08277cs.CLcs.LG
收藏

面向基于证据的问答的忠实且鲁棒的LLM专家

Towards Faithful and Robust LLM Specialists for Evidence-Based Question-Answering

University of Zürich(苏黎世大学);ETH Zürich(苏黎世联邦理工学院);University of Regensburg(雷根斯堡大学);Swiss Finance Institute (SFI)(瑞士金融学院)

Tobias Schimanski, Jingwei Ni, Mathias Kraus, Elliott Ash, Markus Leippold

AI总结:

本文研究如何通过带有自动化质量过滤器的数据生成流程对LLMs进行鲁棒微调,以提升基于证据问答的来源质量和答案可归因性,并证明数据质量比数量更重要。

摘要:

大型语言模型(LLMs)在提供更忠实和可追溯的答案方面的进展,对各种研究和实际应用至关重要。实现这一目标的途径之一是将答案建立在可靠来源的基础上。然而,事实证明,这种基于证据的问答(Evidence-Based QA)在LLMs中效果不佳,主要体现在引用正确的来源(来源质量)以及真实地呈现来源中的信息(答案可归因性)方面。在本研究中,我们系统地探讨了如何对LLMs进行鲁棒的微调,以提升来源质量和答案可归因性。具体而言,我们引入了一个带有自动化数据质量过滤器的数据生成流程,该流程能够大规模合成多样化的高质量训练和测试数据。我们进一步引入了四个测试集,以对微调后的专家模型进行鲁棒性基准测试。广泛的评估表明,在合成数据上进行微调可以提升模型在分布内和分布外的性能。此外,我们证明了数据质量(通过所提出的质量过滤器可以得到显著提升)在改进基于证据的问答中比数量更为重要。

英文摘要:

Advances towards more faithful and traceable answers of Large Language Models (LLMs) are crucial for various research and practical endeavors. One avenue in reaching this goal is basing the answers on reliable sources. However, this Evidence-Based QA has proven to work insufficiently with LLMs in terms of citing the correct sources (source quality) and truthfully representing the information within sources (answer attributability). In this work, we systematically investigate how to robustly fine-tune LLMs for better source quality and answer attributability. Specifically, we introduce a data generation pipeline with automated data quality filters, which can synthesize diversified high-quality training and testing data at scale. We further introduce four test sets to benchmark the robustness of fine-tuned specialist models. Extensive evaluation shows that fine-tuning on synthetic data improves performance on both in- and out-of-distribution. Furthermore, we show that data quality, which can be drastically improved by proposed quality filters, matters more than quantity in improving Evidence-Based QA.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑