Xiaomi's Large Model Team Claims Top Spot on MMAU Audio Reasoning Leaderboard
2025-03-17 / Read about 0 minute
Author:小编   

Drawing inspiration from Deepseek-R1, Xiaomi's Large Model Team has successfully integrated reinforcement learning algorithms into audio understanding. This groundbreaking approach has led to an impressive top score of 64.5% accuracy on the esteemed international MMAU audio understanding evaluation benchmark. Furthermore, the team has generously made the code open source, fostering collaboration and advancement in the field.