Am I Me or You? State-of-the-Art Dialogue Models Cannot Maintain an Identity
This addresses a critical issue for dialogue systems in maintaining factual accuracy and consistency, though it remains a challenging problem.
The paper tackled the problem of dialogue models failing to maintain character identity, such as taking on the interlocutor's role, and showed that their best models reduced mistaken identity issues by nearly 65% while improving engagingness.
State-of-the-art dialogue models still often stumble with regards to factual accuracy and self-contradiction. Anecdotally, they have been observed to fail to maintain character identity throughout discourse; and more specifically, may take on the role of their interlocutor. In this work we formalize and quantify this deficiency, and show experimentally through human evaluations that this is indeed a problem. In contrast, we show that discriminative models trained specifically to recognize who is speaking can perform well; and further, these can be used as automated metrics. Finally, we evaluate a wide variety of mitigation methods, including changes to model architecture, training protocol, and decoding strategy. Our best models reduce mistaken identity issues by nearly 65% according to human annotators, while simultaneously improving engagingness. Despite these results, we find that maintaining character identity still remains a challenging problem.