# Scaling Up Music Information Retrieval Training with Semi-Supervised Learning

arXiv: https://arxiv.org/abs/2310.01353 · HTML: https://mlsanity.com/paper/2310.01353/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Speech and audio distillation, one of 12 directions in Knowledge distillation.
- Builds on 17 of its 37 references that are in the map.
- First posted on arXiv 2 Oct 2023.

## Placed in

- Knowledge distillation / [Speech and audio distillation](https://mlsanity.com/distill/speech-and-audio-distillation/)

## Builds on (most cited first, 17 of 17)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Identity Mappings in Deep Residual Networks](https://mlsanity.com/paper/1603.05027/) (2016)
- [Self-training with Noisy Student improves ImageNet classification](https://mlsanity.com/paper/1911.04252/) (2019)
- [Perceiver: General Perception with Iterative Attention](https://mlsanity.com/paper/2103.03206/) (2021)
- [Jukebox: A Generative Model for Music](https://mlsanity.com/paper/2005.00341/) (2020)
- [Self-supervised Pretraining of Visual Features in the Wild](https://mlsanity.com/paper/2103.01988/) (2021)
- [MERT: Acoustic Music Understanding Model with Large-Scale Self-supervised Training](https://mlsanity.com/paper/2306.00107/) (2023)
- [Codified audio language modeling learns useful representations for music information retrieval](https://mlsanity.com/paper/2107.05677/) (2021)
- [Semi-Supervised Music Tagging Transformer](https://mlsanity.com/paper/2111.13457/) (2021)
- [Melody transcription via generative pre-training](https://mlsanity.com/paper/2212.01884/) (2022)
- [SpecTNT: a Time-Frequency Transformer for Music Audio](https://mlsanity.com/paper/2110.09127/) (2021)
- [Scaling ASR Improves Zero and Few Shot Learning](https://mlsanity.com/paper/2111.05948/) (2021)
- [End-to-End Musical Key Estimation Using a Convolutional Neural Network](https://mlsanity.com/paper/1706.02921/) (2017)
- [Multitrack Music Transcription with a Time-Frequency Perceiver](https://mlsanity.com/paper/2306.10785/) (2023)
- [Modeling Beats and Downbeats with a Time-Frequency Transformer](https://mlsanity.com/paper/2205.14701/) (2022)
- [To catch a chorus, verse, intro, or anything else: Analyzing a song with structural functions](https://mlsanity.com/paper/2205.14700/) (2022)
- [MuSFA: Improving Music Structural Function Analysis with Partially Labeled Data](https://mlsanity.com/paper/2211.15787/) (2022)

## Built on by (most cited first, 2 of 2)

- [Transcribing Rhythmic Patterns of the Guitar Track in Polyphonic Music](https://mlsanity.com/paper/2510.05756/) (2025)
- [Enhancing Automatic Chord Recognition via Pseudo-Labeling and Knowledge Distillation](https://mlsanity.com/paper/2602.19778/) (2026)

## Abstract (arXiv)

In the era of data-driven Music Information Retrieval (MIR), the scarcity of labeled data has been one of the major concerns to the success of an MIR task. In this work, we leverage the semi-supervised teacher-student training approach to improve MIR tasks. For training, we scale up the unlabeled music data to 240k hours, which is much larger than any public MIR datasets. We iteratively create and refine the pseudo-labels in the noisy teacher-student training process. Knowledge expansion is also explored to iteratively scale up the model sizes from as small as less than 3M to almost 100M parameters. We study the performance correlation between data size and model size in the experiments. By scaling up both model size and training data, our models achieve state-of-the-art results on several MIR tasks compared to models that are either trained in a supervised manner or based on a self-supervised pretrained model. To our knowledge, this is the first attempt to study the effects of scaling up both model and training data for a variety of MIR tasks.