• [paper link]
• Abstract
• Each sample is presented along with the top-K probes’ monitored decision along the time-axis. The yellow line, green line, and the shaded blue region denote the median, mean, and standard deviation of inferred outputs by the probes, respectively. The red dashed line indicates the threshold value $\tau$. A horizontal gray dashed line marks 0.5 probability, and a vertical gray dashed line is presented upon Audio Continuation approach to segregate the input from the generated music.
• We plan to open-source the code soon.

all audio samples are generated with the same random seed, and normalized to -12 dB LUFS

Samples - 3.3. Qualitative Evaluation

Samples - 3.4. Objective & 3.5. Subjective Evaluation

Samples - 4.1. Multiple Direction

Samples - 4.2. Generating “Realistic” Music

Samples - 4.3. Effects on Number of Probing Data

Bonus

Failing Cases

Other ITI task: Removing instruments

Other ITI task: “more cowbell”