CoDi-2 vs. AUDIT ? #136
Description
Thank you to the authors for their excellent work. However, I have some concerns. In the evaluation of the audio editing task, the authors compared their results with the publicly reported metrics from the AUDIT paper. However, this comparison may not be entirely fair, as the dataset used in this work is AudioSet, while the dataset used in the original AUDIT paper differs from the one used here. Could the authors clarify whether such a comparison is meaningful under these circumstances? Have they considered the potential implications of this discrepancy?