arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2404.06209cs.LGcs.AIcs.CL
收藏

大象永不遗忘:大型语言模型中表格数据的记忆与学习

Elephants Never Forget: Memorization and Learning of Tabular Data in Large Language Models

University of Tübingen(蒂宾根大学);Microsoft Research(微软研究院)

Sebastian Bordt, Harsha Nori, Vanessa Rodrigues, Besmira Nushi, Rich Caruana

AI总结:

本文提出多种检测方法评估 LLM 是否记忆表格数据集,发现其逐字记忆常见数据并导致性能虚高,同时揭示少样本学习样本效率不足,并发布 Python 检测工具。

摘要:

尽管许多研究已经展示了大型语言模型(LLM)如何能够应用于多种多样的任务,但数据污染和记忆这两个关键问题往往被轻描淡写。在本工作中,我们针对表格数据处理这一关切。具体而言,我们引入了多种不同技术,以评估语言模型在训练期间是否见过某个表格数据集。该调查揭示,LLM 已经逐字记忆了许多流行的表格数据集。随后,我们将 LLM 在训练期间见过的数据集上的少样本学习性能,与其在训练之后发布的数据集上的性能进行比较。我们发现,LLM 在训练期间见过的数据集上表现更好,这表明记忆会导致过拟合。与此同时,LLM 在新颖数据集上表现出不可忽视的性能,并且对数据变换具有出人意料的鲁棒性。接着,我们研究 LLM 的上下文统计学习能力。尽管 LLM 在解决统计分类问题时显著优于随机水平,但少样本学习的样本效率落后于传统统计学习算法,尤其是在问题维度增加时。这表明,在新颖真实世界数据集上观察到的少样本性能,很大一部分来自 LLM 的世界知识。总体而言,我们的结果强调了检验 LLM 在预训练期间是否见过评估数据集的重要性。我们发布了 https://github.com/interpretml/LLM-Tabular-Memorization-Checker Python 包,用于测试 LLM 对表格数据集的记忆情况。

英文摘要:

While many have shown how Large Language Models (LLMs) can be applied to a diverse set of tasks, the critical issues of data contamination and memorization are often glossed over. In this work, we address this concern for tabular data. Specifically, we introduce a variety of different techniques to assess whether a language model has seen a tabular dataset during training. This investigation reveals that LLMs have memorized many popular tabular datasets verbatim. We then compare the few-shot learning performance of LLMs on datasets that were seen during training to the performance on datasets released after training. We find that LLMs perform better on datasets seen during training, indicating that memorization leads to overfitting. At the same time, LLMs show non-trivial performance on novel datasets and are surprisingly robust to data transformations. We then investigate the in-context statistical learning abilities of LLMs. While LLMs are significantly better than random at solving statistical classification problems, the sample efficiency of few-shot learning lags behind traditional statistical learning algorithms, especially as the dimension of the problem increases. This suggests that much of the observed few-shot performance on novel real-world datasets is due to the LLM's world knowledge. Overall, our results highlight the importance of testing whether an LLM has seen an evaluation dataset during pre-training. We release the https://github.com/interpretml/LLM-Tabular-Memorization-Checker Python package to test LLMs for memorization of tabular datasets.

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。

↑