A Recipe For Building a Compliant Real Estate Chatbot

TL;DR

This work focuses on developing a chatbot specialized in the real estate domain, with an emphasis on incorporating compliant behavior to ensure it can be used without perpetuating discriminatory practices, which have historically plagued the real estate industry in the United States.

Abstract

In recent years, there has been significant effort to align large language models with human preferences. This work focuses on developing a chatbot specialized in the real estate domain, with an emphasis on incorporating compliant behavior to ensure it can be used without perpetuating discriminatory practices like steering and redlining, which have historically plagued the real estate industry in the United States. Building on prior work, we present a method for generating a synthetic general instruction-following dataset, along with safety data. Through extensive evaluations and benchmarks, we fine-tuned a llama-3-8B-instruct model and demonstrated that we can enhance it's performance significantly to match huge closed-source models like GPT-4o while making it safer and more compliant. We open-source the model, data and code to support further development and research in the community.

Figures (19)

• Figure 1: An example of non-compliant behavior of GPT-4o as a real estate chatbot compared with our proposed model.
• Figure 2: General synthetic instruction following dataset creation pipeline. Note that we are showing an instance of the generated prompt.
• Figure 3: Performance of different models on the four proposed G-Eval metrics
• Figure 4: Pairwise head-to-head win rate of the models on the four metrics. Note that there is a threshold of 1% for ties to highlight more significant differences. The cells denote the win rate of left models vs the top models.
• Figure 5: 15 Most frequent topics along with their 5 most frequent subtopics for the general instructions split of the data