Prompted LLMs as Chatbot Modules for Long Open-domain Conversation

TL;DR

This paper introduces MPC, a modular prompting approach that leverages pre-trained large language models as chatbot components, enabling high-quality, consistent, and flexible long-term open-domain conversations without fine-tuning.

Contribution

It presents a novel modular prompting framework for chatbots that maintains performance comparable to fine-tuned models using only prompting techniques.

Findings

MPC achieves human-evaluated quality comparable to fine-tuned chatbots.

The approach enhances long-term consistency and flexibility in open-domain conversations.

Human evaluations confirm the effectiveness of MPC in real-world settings.

Abstract

In this paper, we propose MPC (Modular Prompted Chatbot), a new approach for creating high-quality conversational agents without the need for fine-tuning. Our method utilizes pre-trained large language models (LLMs) as individual modules for long-term consistency and flexibility, by using techniques such as few-shot prompting, chain-of-thought (CoT), and external memory. Our human evaluation results show that MPC is on par with fine-tuned chatbot models in open-domain conversations, making it an effective solution for creating consistent and engaging chatbots.