Large Language Models Engineer Too Many Simple Features for Tabular Data
Keywords: LLM, Feature Engineering, Bias, Tabular Data
TL;DR: LLMs are biased towards simple operators when performing feature engineering for tabular data which hinders their ability to generate quality features.
Abstract: Tabular machine learning problems often require time-consuming and labor-intensive feature engineering. Recent efforts have focused on using large language models (LLMs) to capitalize on their potential domain knowledge. At the same time, researchers have observed ethically concerning negative biases in other LLM-related use cases, such as text generation. These developments motivated us to investigate whether LLMs exhibit a bias that negatively impacts the performance of feature engineering. While not ethically concerning, such a bias could hinder practitioners from fully utilizing LLMs for automated data science. Therefore, we propose a method to detect potential biases by detecting anomalies in the frequency of operators (e.g., adding two features) suggested by LLMs when engineering new features. We detect anomalies by comparing the frequencies of operators from LLMs to those of the optimal operators. Our experiments evaluate the bias of four LLMs, two big frontier and two small open-source models, across 27 tabular datasets. Our results indicate that LLMs are biased toward simple operators, such as addition, and can fail to utilize more complex operators, such as grouping followed by aggregations. Furthermore, the bias can negatively impact the predictive performance when using LLM-generated features. Our results call for mitigating bias when feature engineering with LLMs.
Submission Number: 51
Loading