MuTual: A Dataset for Multi-Turn Dialogue Reasoning
This dataset facilitates research into improving reasoning abilities in non-task oriented dialogue systems, addressing a specific bottleneck in AI conversation models.
The authors introduced MuTual, a dataset of 8,860 annotated dialogues for multi-turn dialogue reasoning, based on Chinese student English listening exams, to address logical mistakes in non-task oriented dialogue systems. Empirical results show state-of-the-art methods achieve only 71% accuracy, far behind human performance of 94%, highlighting a significant gap in reasoning capabilities.
Non-task oriented dialogue systems have achieved great success in recent years due to largely accessible conversation data and the development of deep learning techniques. Given a context, current systems are able to yield a relevant and fluent response, but sometimes make logical mistakes because of weak reasoning capabilities. To facilitate the conversation reasoning research, we introduce MuTual, a novel dataset for Multi-Turn dialogue Reasoning, consisting of 8,860 manually annotated dialogues based on Chinese student English listening comprehension exams. Compared to previous benchmarks for non-task oriented dialogue systems, MuTual is much more challenging since it requires a model that can handle various reasoning problems. Empirical results show that state-of-the-art methods only reach 71%, which is far behind the human performance of 94%, indicating that there is ample room for improving reasoning ability. MuTual is available at https://github.com/Nealcly/MuTual.