arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2411.02671cs.LGcs.AIcs.CL
收藏

通过潜在概念变量实现公平的上下文学习

Fair In-Context Learning via Latent Concept Variables

University of Arkansas(亚拉巴马大学);Baylor University(贝勒大学);University of Texas at Dallas(德克萨斯大学达拉斯分校)

Karuna Bhaila, Minh-Hao Van, Kennedy Edemacu, Chen Zhao, Feng Chen, Xintao Wu

AI总结:

针对LLMs在表格数据上下文学习中的固有偏见问题,提出利用潜在概念变量的最优演示选择方法,通过数据增强降低预测与敏感变量的相关性,经内部LLM学习概念并泛化至外部LLMs,提升了表格数据集上的公平性结果。

摘要:

大型语言模型(LLMs)新兴的上下文学习(ICL)能力促使其通过序列化方法在不同数据类型(包括表格数据)的各类领域中用于预测任务。然而,随着在高风险领域应用的增加,研究表明LLMs可能从预训练数据中继承社会偏见和歧视。本研究探讨LLMs在表格数据上下文学习过程中的固有偏见,聚焦于利用潜在概念变量实现资源高效任务适配的最优演示选择方法。我们设计了数据增强策略,以降低预测结果与敏感变量的相关性,助力在潜在概念学习中提升公平性;利用所学概念选择演示并获得公平预测。潜在概念变量通过较小的内部LLM学习,并泛化至较大的外部LLMs。实证验证表明,与多种启发式演示选择方法相比,该公平潜在变量方法在表格数据集上提升了公平性结果。

英文摘要:

The emerging in-context learning (ICL) ability of large language models (LLMs) has prompted their use for predictive tasks in various domains with different data types, including tabular data, facilitated by serialization methods. However, with increasing applications in high-stakes domains, it has been shown that LLMs can inherit social bias and discrimination from their pre-training data. In this work, we investigate inherent bias in LLMs during in-context learning with tabular data. We focus on an optimal demonstration selection approach that utilizes latent concept variables for resource-efficient task adaptation. We design data augmentation strategies that reduce the correlation between predictive outcomes and sensitive variables, helping promote fairness during latent concept learning. We utilize the learned concept to select demonstrations and obtain fair predictions. The latent concept variables are learned using a smaller internal LLM and generalized to larger external LLMs. We empirically verify that the fair latent variable approach improves fairness results on tabular datasets compared to multiple heuristic demonstration selection methods.

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。

↑