Open-Unmix Music Separation Quiz

Answers and explanations appear as you take the quiz.

1. What does Open-Unmix do?
   • Separates a song into stems like vocals, drums, bass, and other
   • Generates lyrics from melodies
   • Compresses audio into MP3
   • Tunes guitars automatically
2. What neural network is at the core of Open-Unmix?
   • A convolutional GAN
   • A three-layer bidirectional LSTM with dense layers
   • A decision forest
   • A diffusion transformer
3. On what representation does Open-Unmix operate?
   • The raw waveform only
   • MIDI note events
   • The magnitude spectrogram of the mixture
   • Sheet music images
4. Which dataset is Open-Unmix trained on?
   • ImageNet
   • LibriSpeech
   • COCO
   • MUSDB18
5. What was the main design goal of Open-Unmix?
   • To be a transparent, reproducible reference baseline
   • To win every leaderboard at any cost
   • To run only on proprietary hardware
   • To replace human musicians
6. How does Open-Unmix reconstruct a separated stem after predicting a mask?
   • It synthesizes audio from scratch with a vocoder
   • It multiplies the mask by the mixture magnitude and reuses the original phase via inverse STFT
   • It plays back the original file unchanged
   • It converts the mask to text
7. In the original Open-Unmix reference design, how are target stems modeled?
   • One language model predicts all lyrics
   • A single fixed equalizer separates every instrument
   • A separate target-specific network is trained for each stem
   • A classifier only labels the whole song
8. What is a stated reason Open-Unmix uses framework-specific implementations?
   • To prevent researchers from comparing results
   • To require proprietary hardware
   • To eliminate all pre- and post-processing
   • To keep implementations clear and reproducible across supported frameworks