# Accept or Deny? Evaluating LLM Fairness and Performance in Loan Approval across Table-to-Text Serialization Approaches

## TL;DR

This study evaluates how different serialization formats of tabular data impact the fairness and performance of LLMs in loan approval tasks across diverse regions, highlighting the importance of data representation methods.

## Contribution

It demonstrates that serialization formats significantly influence LLM fairness and accuracy, and assesses the effects of in-context learning on these metrics in high-stakes decision-making.

## Key findings

- Serialization format affects LLM performance and fairness.
- In-context learning improves accuracy but varies in fairness impact.
- Certain formats like GReat and LIFT increase F1 scores but worsen fairness disparities.

## Abstract

Large Language Models (LLMs) are increasingly employed in high-stakes decision-making tasks, such as loan approvals. While their applications expand across domains, LLMs struggle to process tabular data, ensuring fairness and delivering reliable predictions. In this work, we assess the performance and fairness of LLMs on serialized loan approval datasets from three geographically distinct regions: Ghana, Germany, and the United States. Our evaluation focuses on the model's zero-shot and in-context learning (ICL) capabilities. Our results reveal that the choice of serialization (Serialization refers to the process of converting tabular data into text formats suitable for processing by LLMs.) format significantly affects both performance and fairness in LLMs, with certain formats such as GReat and LIFT yielding higher F1 scores but exacerbating fairness disparities. Notably, while ICL improved model performance by 4.9-59.6% relative to zero-shot baselines, its effect on fairness varied considerably across datasets. Our work underscores the importance of effective tabular data representation methods and fairness-aware models to improve the reliability of LLMs in financial decision-making.