V-VAE: A Variational Auto Encoding Framework Towards Fine-Grained Control over Human-Like Chat

TL;DR

This paper introduces V-VAE, a novel variational auto-encoding framework that enables fine-grained, dynamic control over human-like chat responses by modeling subtle latent traits, supported by a new high-quality dataset and benchmark.

Contribution

V-VAE provides a new method for modeling dynamic, fine-grained traits in human-like chat, overcoming limitations of static role descriptions and synthetic data.

Findings

V-VAE outperforms standard baselines on HumanChatBench and DialogBench.

The framework effectively captures subtle latent traits such as emotional tone and personality.

The high-quality HumanChatData dataset enhances training and evaluation of human-like chat models.

Abstract

With the continued proliferation of Large Language Model (LLM) based chatbots, there is a growing demand for generating responses that are not only linguistically fluent but also consistently aligned with persona-specific traits in conversations. However, existing role-play and persona-based chat approaches rely heavily on static role descriptions, coarse-grained signal space, and low-quality synthetic data, which fail to capture dynamic fine-grained details in human-like chat. Human-like chat requires modeling subtle latent traits, such as emotional tone, situational awareness, and evolving personality, which are difficult to predefine and cannot be easily learned from synthetic or distillation-based data. To address these limitations, we propose a Verbal Variational Auto-Encoding (V-VAE) framework, containing a variational auto-encoding module and fine-grained control space which dynamically adapts dialogue behaviour based on fine-grained, interpretable latent variables across talking style, interaction patterns, and personal attributes. We also construct a high-quality dataset, HumanChatData, and benchmark HumanChatBench to address the scarcity of high-quality data in the human-like domain. Experiments show that LLMs based on V-VAE consistently outperform standard baselines on HumanChatBench and DialogBench, which further demonstrates the effectiveness of V-VAE and HumanChatData.