MiDashengLM: Efficient Audio Understanding with General Audio Captions

Benchmark Model Rank Results
audio-classification-on-vggsoundMiDashengLM-7B#14Top 1 Accuracy: 52.1
audio-tagging-on-audiosetMiDashengLM-7B#18mean average precision: 0.089
speech-recognition-on-librispeech-test-otherMiDashengLM-7B#111Word Error Rate (WER): 6.2
speech-recognition-on-librispeech-test-cleanMiDashengLM-7B#126Word Error Rate (WER): 3.7