Training Dialogue Systems by AI Feedback for Improving Overall Dialogue Impression
This work addresses the problem of enhancing user engagement in AI dialogue systems by focusing on holistic dialogue impressions, representing an incremental improvement over existing reinforcement learning from AI feedback methods.
The study tackled the challenge of improving overall dialogue impressions like consistency and empathy in dialogue systems by developing a reward model based on supervised fine-tuning of LLMs for 12 metrics, and tuning dialogue models with this feedback, resulting in improved automatic and human evaluations of metrics and response naturalness.
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.