@inproceedings{LLMembedding,
author = {Li, Kang and Miao, Jiawei and Cucuringu, Mihai and S\'{a}nchez-Betancourt, Leandro},
title = {LLM Embedding for Regression Priors},
year = {2025},
isbn = {9798400722202},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3768292.3770437},
doi = {10.1145/3768292.3770437},
abstract = {We propose Semantic Prior Regression (SPR), a novel Bayesian framework that systematically incorporates semantic knowledge from large language model embeddings into regression priors. The methodology transforms natural language descriptions of response and predictor variables into high-dimensional embeddings, quantifies semantic relatedness through cosine similarity metrics, and constructs informative priors based on these similarities. The framework combines semantic priors with numerical likelihood through standard Bayesian conjugacy, enabling closed-form posterior inference. We demonstrate applications to macroeconomic forecasting using Federal Reserve Economic Data (FRED). Results show that semantic priors significantly enhance predictive accuracy in data-scarce environments while maintaining interpretability. The approach bridges natural language understanding and statistical inference, offering a principled method for incorporating domain knowledge into regression analysis.},
booktitle = {Proceedings of the 6th ACM International Conference on AI in Finance},
pages = {220-228},
numpages = {9},
keywords = {Bayesian regression, large language models, semantic embeddings, prior construction, financial forecasting},
location = {
},
series = {ICAIF '25}
}