NUS's DuET: Dual-Encoder Transformer Jointly Does MIDI-to-Score Transcription and Beat Tracking
NationalUniversityofSingapore · hf · 2026-10-19
NUS researchers introduce DuET, a note-frame dual-encoder transformer that jointly performs MIDI-to-score transcription and beat tracking for piano performances.
- Architecture: 8 dual-encoder layers plus 4 alternating decoder pairs, with separate note and frame streams using TPRoPE/TimeRoPE positional encodings, plus a beat/downbeat head
- Audio is first transcribed to performance MIDI via Transkun v2; output exports to MusicXML viewable in MuseScore 4
- Qualitative comparisons show DuET produces cleaner barlines and metrical structure than Beyer & Dai and neural beat tracking baselines
More from Research
- Domingos Cites New Result Claiming RL Is 'Basically Useless' vs Supervised Learning — pmddomingos · 2026-10-23
- Wolfram Opens Rule 30 Prizes to AIs: $30,000 for Three Open Problems — cephaloform · 2026-10-23
- Stanford and Databricks use AI agents to reproduce medical findings and test analytical robustness — ChengleiSi · 2026-10-23
- EMNLP paper: diffusion LMs internally encode latent denoising timestep representations — gsarti_ · 2026-10-23
- KU Leuven and VIB open two Group Leader professorships in AI for Biology, start 2027 — jpillowtime · 2026-10-23
- Subspace Tuning curbs LLM catastrophic forgetting by freezing principal singular directions — scottniekum · 2026-10-23