DynDST: A Dynamic Dialogue State Tracking Dataset for Assessing the Conversational Adaptability of Large Language Models

Anonymous

Abstract: This work tackles a key challenge in dialogue systems: the ability to adapt to changing user intentions and resolve inconsistencies in conversation histories. This is crucial in scenarios like train ticket booking, where customer plans often change dynamically. Despite advancements in NLP and large language models (LLMs), these systems struggle with real-time information updates during conversations.We introduce a specialized dataset to evaluate chatbot models on dynamic dialogue state tracking, focusing on scenarios where users modify their requests mid-conversation. This work aims to improve chatbot coherence and consistency, bridging the gap between the current capabilities of dialogue systems and the fluidity of human-like conversational interactions.

Paper Type: short

Research Area: Discourse and Pragmatics

Contribution Types: Data resources

Languages Studied: English