1 paper
Kyudan Jung, Sehyun Lee, Song-ha Jo +4
Continuous audio encoders compress audio along feature and time axes through latent width and frame rate, but their joint effect on downstream performance remains unclear. We train…