distributional drift

The paper addresses the issue of distributional shift in machine learning models, especially in text-to-audio generation. The authors observe a consistent audio quality degradation in generated audio samples with user…