Training Dialogue Systems by AI Feedback for Improving Overall Dialogue Impression

TL;DR

This paper introduces a supervised fine-tuning approach using reward models based on LLMs to enhance dialogue system impressions like consistency and empathy, showing improved responses through automatic and human evaluations.

Contribution

It proposes a novel supervised fine-tuning method with reward models for evaluating and improving overall dialogue impressions, addressing challenges in dialogue-level evaluation.

Findings

Improved dialogue response naturalness after fine-tuning

Enhanced evaluation metrics for dialogue impressions

Better alignment with human judgments

Abstract

To improve user engagement during conversations with dialogue systems, we must improve individual dialogue responses and dialogue impressions such as consistency, personality, and empathy throughout the entire dialogue. While such dialogue systems have been developing rapidly with the help of large language models (LLMs), reinforcement learning from AI feedback (RLAIF) has attracted attention to align LLM-based dialogue models for such dialogue impressions. In RLAIF, a reward model based on another LLM is used to create a training signal for an LLM-based dialogue model using zero-shot/few-shot prompting techniques. However, evaluating an entire dialogue only by prompting LLMs is challenging. In this study, the supervised fine-tuning (SFT) of LLMs prepared reward models corresponding to 12 metrics related to the impression of the entire dialogue for evaluating dialogue responses. We tuned our dialogue models using the reward model signals as feedback to improve the impression of the system. The results of automatic and human evaluations showed that tuning the dialogue model using our reward model corresponding to dialogue impression improved the evaluation of individual metrics and the naturalness of the dialogue response.