Researchers address the critical shortage of meeting data by creating FAME, a dataset of synthetic meeting transcripts generated through multi-agent simulation.

• Developed MIMIC, a framework that orchestrates LLM-powered agents with distinct psychological profiles to debate given topics
• Generated 500 English and 300 German synthetic meeting transcripts
• Demonstrated that these synthetic conversations effectively train summarization models

This breakthrough enables organizations to develop and test meeting summarization tools without privacy concerns or expensive data collection, potentially accelerating the development of AI assistants for workplace meetings.

Original Paper: You need to MIMIC to get FAME: Solving Meeting Transcript Scarcity with a Multi-Agent Conversations