arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2403.17860cs.CL
收藏

用目标智能体在环合成数据照亮语言模型的盲区

Illuminating Blind Spots of Language Models with Targeted Agent-in-the-Loop Synthetic Data

Delft University of Technology(代尔夫特理工大学)

Philip Lippmann, Matthijs T. J. Spaan, Jie Yang

AI总结:

本研究提出利用人类或大型语言模型作为智能代理教师,通过生成目标合成数据来减少语言模型的高置信度错误分类(未知的未知),在三个分类任务上验证了有效性,并发现语言模型在成本效益上优于人类。

摘要:

语言模型(LMs)在各种任务上取得了令人印象深刻的准确性,但仍然容易受到高置信度错误分类的影响,这也被称为未知的未知(UUs)。这些UUs在特征空间中聚集成盲区,导致在高风险应用中存在显著风险。这对于更小、更轻量级的语言模型尤其相关,因为它们更容易受到此类错误的影响。虽然UUs的识别已被广泛研究,但其缓解仍然是一个开放的挑战,包括如何使用已识别的UUs来消除未见过的盲区。在这项工作中,我们提出了一种新颖的方法,通过使用智能代理——无论是人类还是大型语言模型——作为教师来表征UU型错误,从而解决盲区缓解问题。通过利用智能代理的泛化能力,我们识别高置信度错误分类中的模式,并使用它们生成有针对性的合成样本,以提高模型鲁棒性并减少盲区。我们在三个分类任务上对我们的方法进行了广泛评估,并证明了其在减少UUs数量方面的有效性,同时保持了相似的准确性水平。我们发现,人类计算的有效性具有很高的上限,但高度依赖于对底层任务的熟悉程度。此外,人类与语言模型之间的成本差距超过一个数量级,因为语言模型在更具可扩展性的同时,达到了类似人类的泛化和生成性能。

英文摘要:

Language models (LMs) have achieved impressive accuracy across a variety of tasks but remain vulnerable to high-confidence misclassifications, also referred to as unknown unknowns (UUs). These UUs cluster into blind spots in the feature space, leading to significant risks in high-stakes applications. This is particularly relevant for smaller, lightweight LMs that are more susceptible to such errors. While the identification of UUs has been extensively studied, their mitigation remains an open challenge, including how to use identified UUs to eliminate unseen blind spots. In this work, we propose a novel approach to address blind spot mitigation through the use of intelligent agents -- either humans or large LMs -- as teachers to characterize UU-type errors. By leveraging the generalization capabilities of intelligent agents, we identify patterns in high-confidence misclassifications and use them to generate targeted synthetic samples to improve model robustness and reduce blind spots. We conduct an extensive evaluation of our method on three classification tasks and demonstrate its effectiveness in reducing the number of UUs, all while maintaining a similar level of accuracy. We find that the effectiveness of human computation has a high ceiling but is highly dependent on familiarity with the underlying task. Moreover, the cost gap between humans and LMs surpasses an order of magnitude, as LMs attain human-like generalization and generation performance while being more scalable.

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。

↑