With the rise of large language models (LLMs) for flexibly processing information as strings, a natural application is regression, specifically by preprocessing string representations into LLM embeddings as downstream features for metric prediction. In this paper, we provide one of the first comprehensive investigations into embedding-based regression and demonstrate that LLM embeddings as features can be better for high-dimensional regression tasks than using traditional feature engineering. This regression performance can be explained in part due to LLM embeddings over numeric data inherently preserving Lipschitz continuity over the feature space. Furthermore, we quantify the contribution of different model effects, most notably model size and language understanding, which we find surprisingly do not always improve regression performance.
@article{arxiv.2411.14708,
title = {Understanding LLM Embeddings for Regression},
author = {Eric Tang and Bangding Yang and Xingyou Song},
journal= {arXiv preprint arXiv:2411.14708},
year = {2025}
}
Comments
Published in Transactions on Machine Learning Research (TMLR) 2025. Code can be found in https://github.com/google-research/optformer