PlatoLM: Teaching LLMs in Multi-Round Dialogue via a User Simulator

TL;DR

PlatoLM is a new response model trained with human-like questions from genuine conversations, achieving state-of-the-art performance and more human-like dialogue qualities in multi-round interactions.

Contribution

We introduce a novel user simulator 'Socratic' that enhances multi-round dialogue training by incorporating genuine human questions, improving model human-likeness and topic diversity.

Findings

PlatoLM outperforms existing LLaMA-based models on MT-Bench.

Our method produces more human-like questioning patterns.

The approach enriches topic diversity in multi-turn conversations.

Abstract

The unparalleled performance of closed-sourced ChatGPT has sparked efforts towards its democratization, with notable strides made by leveraging real user and ChatGPT dialogues, as evidenced by Vicuna. However, due to challenges in gathering dialogues involving human participation, current endeavors like Baize and UltraChat rely on ChatGPT conducting roleplay to simulate humans based on instructions, resulting in overdependence on seeds, diminished human-likeness, limited topic diversity, and an absence of genuine multi-round conversational dynamics. To address the above issues, we propose a paradigm to simulate human behavior better and explore the benefits of incorporating more human-like questions in multi-turn conversations. Specifically, we directly target human questions extracted from genuine human-machine conversations as a learning goal and provide a novel user simulator called `Socratic'. The experimental results show our response model, `PlatoLM', achieves SoTA performance among LLaMA-based 7B models in MT-Bench. Our findings further demonstrate that our method introduces highly human-like questioning patterns and rich topic structures, which can teach the response model better than previous works in multi-round conversations.