Yuelin Hu, Jun Xu, Bingcong Lu, Zhengxue Cheng, Hongwei Hu, Ronghua Wu, Li Song
MeetBench-XL introduces a comprehensive evaluation framework and a dual-policy AI agent for enhancing real-time meeting assistance in enterprise environments.
In today's fast-paced work environments, AI assistants need to handle various tasks during meetings, like quickly finding facts or analyzing discussions for strategic insights. Existing tools often fall short because they don't mimic real-world meeting scenarios well. This study introduces a new dataset and evaluation method to better reflect the complexity of enterprise meetings. It also presents a smart AI agent that can switch between quick and detailed processing to efficiently handle meeting queries, improving over existing systems in terms of speed and accuracy.
