<ul><li>Training conversational question-answering systems with in-domain data is challenging due to its scarcity.</li><li>Traditional top-down methods use a large language model to generate multi-turn dialogues, but lack content control and are susceptible to hallucinations.</li><li>A bottom-up approach is introduced, generating QA pairs first and then combining them into coherent dialogues, offering greater control and precision.</li><li>Human and automated evaluations show that the bottom-up approach produces more realistic and higher-quality dialogues compared to top-down methods.</li></ul>

Bottom-Up Synthesis of Knowledge-Grounded Task-Oriented Dialogues with Iteratively Self-Refined Prompts

Discover more