Pith. sign in

REVIEW 3 major objections 7 minor 40 references

Layer-wise Investigation of Large-Scale Self-Supervised Music Representation Models

T0 review · 3 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Self-supervised music models outperform spectral baselines on every tested task

desk verdict Solid layer-wise benchmark for MusicFM and MuQ, but the acoustic-to-semantic evolution claim needs error bars before it can carry weight. read the letter →

arxiv 2505.16306 v1 pith:BJQ3HDAW submitted 2025-05-22 cs.SD cs.AIeess.AS

Yizhi Zhou , Haina Zhu , Hangting Chen This is my paper
classification cs.SDcs.AIeess.AS
keywords self-supervisedlearningmusicinformationretrievallayer-wiseanalysisrepresentationFMMuQprobingcanonicalcorrelation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks what information is stored in modern self-supervised music models and how to use it well. By probing two large music representation models, MusicFM and MuQ, across 14 music information retrieval tasks, it argues that these models consistently beat a handcrafted spectral feature baseline. The layer-by-layer results show a steady shift: lower layers carry acoustic, objective information such as pitch and singer identity, while higher layers carry semantic, subjective information such as genre and emotion. The paper further claims that for most tasks the best single layer sits in the middle of the network, not the final layer, and that picking one good layer often beats averaging all layers with learned weights.

What carries the argument

The analysis rests on three tools. A frozen-feature probing model, a shallow neural network with one 512-unit hidden layer, is trained on features extracted from each encoder layer separately to measure how much task-relevant information each layer holds. PWCCA, a weighted canonical correlation analysis, is used to quantify how similar each layer's representation is to the model's input Mel spectrum. For feature combination, a softmax-weighted sum of all layer outputs is trained alongside the probe, and compared against picking a single best layer.

What would settle it

Run the same 14-task probing on MusicFM and MuQ with, say, five different random seeds per task and layer, and check whether the best-layer rankings stay consistent; if the optimal layer moves by several layers across seeds, the acoustic-to-semantic layer specialization claim is not a stable property of the models.

Watch

Extended reading notes

Core claim

The central discovery is a layered organization of learned music representations. Both MusicFM and MuQ outperform FBANK-plus-Chroma on every one of the 14 tested tasks, by margins that are large for semantic tasks such as genre classification and emotion arousal. As activations pass through the encoder, their correlation with the input Mel spectrum decreases steadily, measured by PWCCA, and task-level probing shows that objective acoustic labels are best decoded from lower layers while subjective semantic labels are best decoded from higher layers. This means the networks gradually re-encode acoustic detail into higher-level musical meaning, and the useful information for a particular task ends up concentrated in a middle band of layers, with the exact best layer task-dependent.

Load-bearing premise

The paper assumes that a single run of probing at each layer gives a stable ranking of how much task-relevant information each layer holds, but it reports no repeated seeds or significance tests, so adjacent-layer differences could be noise.

Editorial extensions

If this is right

  • For downstream music tasks, extracting features from the best-performing middle layer should outperform using the final layer's output in most cases.
  • The category of the task gives a rough guide to layer choice: acoustic labels point to lower layers and semantic labels to higher layers.
  • Since both models show the same trend, the acoustic-to-semantic gradient is likely a shared property of this family of self-supervised music encoders.
  • Learned weighted sums of all layers rarely beat the best single layer on smaller datasets, suggesting that aggregation can hurt when the probe has little data.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Layer-picking as a cheap prior: for a new music task, the paper's placement of the optimal layer could be transferred by task type, saving the cost of scanning all layers.
  • The single-run probing numbers carry no error bars, so small differences between adjacent layers may not be stable; a repeated-seed study could either confirm or dissolve the reported layer rankings.
  • The same layer-wise probing could be applied to the pre-training objective itself, for example testing whether token targets that are more spectral (Mel-RVQ) versus more abstract (random-projection) push the acoustic-to-semantic transition earlier or later.
  • A practical extension would be using the best-layer map to prune models: a task needing only acoustic layers could run a truncated forward pass and save computation.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. The paper presents an empirical, layer-wise analysis of two large self-supervised music representation models, MusicFM and MuQ, on 14 downstream tasks drawn from the MARBLE benchmark. The authors evaluate per-layer probing accuracies, a learned weighted combination of all layers, and PWCCA similarity between layer embeddings and the model input. They report three main findings: (i) both SSL models outperform a FBANK+Chroma baseline on all tasks, often by large margins; (ii) features appear to evolve from acoustic-level to semantic-level information across layers, with acoustic tasks best served by lower layers and semantic tasks by higher layers; and (iii) for most tasks, selecting a single best layer outperforms using a weighted combination of all layers. The paper also observes that the optimal layer is usually not the final layer, and that the two models show a similar qualitative pattern.

Significance. If the results hold, the paper provides practically useful guidance for MIR practitioners: SSL music models are strong feature extractors, middle layers often suffice, and explicit layer selection can be beneficial. The comprehensive comparison across 14 tasks and two model families is a valuable benchmark contribution, and the consistency of the layer-wise trends across MusicFM and MuQ is suggestive of a general property of this model family. The main limitation is that the central interpretative claim—the acoustic-to-semantic progression—rests on single-run probing numbers without error bars or significance tests, and the private MT dataset is not fully specified. The paper would be strengthened by adding statistical validation and clarifying the layer-selection protocol.

major comments (3)
  1. [§4.2, Table 1, Figure 2] The claim that features 'progressively evolve from Acoustic-level to Semantic-level' is inferred from single-run probing accuracies. For MusicFM, the best layers for acoustic tasks are Singer=1, Pitch=1, and Instrument=3, while for semantic tasks they are Genre=5, Emotion=5, Technique=6, and Structure=6; these differences span only a few layers. No confidence intervals, repeated seeds, or significance tests are reported anywhere in the paper, so the observed ordering of best layers could be within run-to-run noise, especially on the small datasets used (GTZAN, Emomusic, VocalSet). In addition, the PWCCA results in Figure 3 show only monotone decreasing similarity to the input, which supports the loss of input-like acoustic information but does not by itself demonstrate that semantic information increases along the proposed acoustic-semantic axis. Please provide variance estimates or significance tests (e.g., paired tests across layers over multiple probe initializations) and, if possible, a more direct measure of semantic content to support the progression claim.
  2. [§3.4, SQC task] The Sound Quality Classification (SQC) task defines the annotation scale as 0 points for bad quality and 1 point for good quality, but then states that 'scores greater than 1 were classified as Good quality while scores of 1 or lower were classified as Bad quality.' Since the maximum possible average score is 1, this threshold would classify every track as Bad, which is inconsistent with the reported accuracies in Table 1 (e.g., 79.5% for MusicFM and 76.5% for the baseline). Please correct the scale or threshold, and provide details on the number of annotators, inter-annotator agreement, and the actual label distribution for all three MT tasks (MPC, RD, SQC).
  3. [§4.3, Table 1] The layer-scanning results are presented with the 'layer where the best result occurred' in parentheses, but the paper does not specify whether this layer is selected on a held-out validation set or on the test set. If the test set is used for layer selection, the comparison against the weighted-sum method is an oracle comparison, and the conclusion that 'directly selecting a single layer produced better results than using all layers' is not a fair evaluation of a practical layer-selection strategy. Please clarify the selection protocol; if validation-based selection was used, report the corresponding test scores, and if not, re-run the comparison with a validation-based selection rule.
minor comments (7)
  1. [Introduction] There is a typo in 'Self-superviesed learning' in the first sentence of the Introduction; it should be 'Self-supervised learning'.
  2. [§3.2] The token sample rate is written as '25K Hz'; this should be '25 kHz' for clarity.
  3. [§3.4] The dataset name 'V ocalSet' contains an extra space; it should be 'VocalSet'.
  4. [§3.1] The sentence 'The training objective is to predict tokens that have been randomly masked within a sequence. similar to BERT.' is fragmentary; please merge the second fragment into the previous sentence.
  5. [§3.4] The 'Comprehensive tasks' category is defined as tasks with 'both semantic and acoustic labels,' yet Key detection is included in this category without a clear justification; please either reclassify Key detection or explain how it fits the definition.
  6. [Figure 2] The red numbers indicating the best-performing layers may be hard to see in grayscale or with color-blind readers; consider adding markers or text labels that do not rely solely on color.
  7. [Related Work] The statement 'To our knowledge, this is the first work to analyze music SSL models on the layer-wise granularity' is a strong claim; please either provide a more thorough search of the literature (including analyses of MERT and other music SSL models) or soften the claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is an external benchmark of frozen SSL checkpoints; the main interpretive claim is under-supported by variance but not by construction.

full rationale

The paper's first claim (SSL models outperform FBANK+Chroma on all 14 tasks) is a direct empirical comparison against external, fixed test sets using frozen checkpoints; no parameter is fitted to force the outcome. The second claim (features evolve from acoustic-level to semantic-level) is an interpretation of best-layer columns in Table 1, Figure 2, and PWCCA curves in Figure 3. The acoustic/semantic task taxonomy is stated up front as an assumption, not derived from the data, and the layer-wise probing numbers are single runs without error bars or significance tests; this makes the interpretive claim fragile, but it does not make it circular, because the data could in principle have shown the opposite ordering. The only self-citation is to MuQ [10], the authors' own model, but MuQ is used as a frozen pretrained checkpoint evaluated on external MARBLE datasets, so the evaluation is independent of any unverified assertion in the MuQ paper. The MT-based tasks use a private dataset, which limits independent reproducibility, but the labels are crowd-sourced annotations and the task definitions are not constructed from model outputs. No equation in the paper reduces a predicted result to an input by definition, and no fitted parameter is renamed as a prediction. Accordingly, no specific circular step can be quoted, and the appropriate score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claim rests on benchmark evaluations and standard interpretability methods; the only hand-set numbers are label thresholds for the private MT dataset.

free parameters (1)
  • MT dataset label thresholds = 0.67, 1.0, 1.34 (Section 3.4)
    Thresholds for mapping average annotator scores to discrete labels (preferred, rap, quality) were chosen after data collection without stability analysis; they affect the MPC, RD, and SQC results.
assumptions (3)
  • domain assumption PWCCA is a valid and robust measure of representation similarity for comparing layers with the input
    The paper relies on PWCCA to claim decreasing similarity from input to higher layers (Figure 3), citing [20-22].
  • domain assumption The categorization of tasks into acoustic, semantic, and comprehensive reflects a meaningful distinction
    The paper uses this categorization to interpret best-layer positions; no inter-annotator agreement or independent validation of the categorization is provided.
  • domain assumption The MARBLE benchmark and the probing setup are standard and unbiased for SSL model comparison
    The paper uses MARBLE tasks and a frozen probing model, following prior work [4], but does not analyze potential biases in these benchmarks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Layer-wise Investigation of Large-Scale Self-Supervised Music Representation Models." pith.science (2026). https://pith.science/paper/BJQ3HDAW

@misc{pith2026250516306,
  author       = {Pith},
  title        = {Pith review of: Layer-wise Investigation of Large-Scale Self-Supervised Music Representation Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BJQ3HDAW}},
  note         = {Machine review of arXiv:2505.16306}
}
read the original abstract

Recently, pre-trained models for music information retrieval based on self-supervised learning (SSL) are becoming popular, showing success in various downstream tasks. However, there is limited research on the specific meanings of the encoded information and their applicability. Exploring these aspects can help us better understand their capabilities and limitations, leading to more effective use in downstream tasks. In this study, we analyze the advanced music representation model MusicFM and the newly emerged SSL model MuQ. We focus on three main aspects: (i) validating the advantages of SSL models across multiple downstream tasks, (ii) exploring the specialization of layer-wise information for different tasks, and (iii) comparing performance differences when selecting specific layers. Through this analysis, we reveal insights into the structure and potential applications of SSL models in music information retrieval.

Figures

Figures reproduced from arXiv: 2505.16306 by the authors.

Figure 1
Figure 1. Model Framework, illustrates the paradigm of using SSL models for downstream tasks. for audio signals. These embeddings represent the features of the music. Then we can use different methods to aggregate the features as the input of downstream models. The entire process is illustrated in the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Layer-wise results on MusicFM and MuQ model, the red numbers indicate the best-performing layers in this column for each task. In addition to these tasks, we define our own semantic-level tasks. We utilized our self-collected MT (Music Taste) dataset for these tasks, which contains 4,000 music samples which are manually annotated. Each sample has three type of different semantic labels. The dataset was then divided … view at source ↗
Figure 3
Figure 3. PWCCA scores between the intermediate layer repre￾sentations and the model input representations task involves frame-level classification. The sample rate of ba￾sic features are equal to the labels, while we need to resample the features of SSL models to match the sample rate. 4.2. How Feature representations evolve We can observe from the layer-wise testing results of MusicFM in [PITH_FULL_IMAGE:figures/full_fig_p… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

40 extracted references · 35 canonical work pages

  1. [1]

    One of the most significant advantages of SSL mod- els is their ability to leverage unlabeled audio data, enabling the possibility of training with large-scale datasets

    Introduction Self-superviesed learning (SSL) has been used to train large fundamental models, including audio signal processing mod- els [1, 2]. One of the most significant advantages of SSL mod- els is their ability to leverage unlabeled audio data, enabling the possibility of training with large-scale datasets. Recently, nu- merous SSL models and method...

    work page

  2. [2]

    SSL models in Music Information Retrieval (MIR) tasks was demonstrated effective through comprehensive experiments

    work page

  3. [3]

    Work performed during an internship at Tencent AI Lab

    Feature representations progressively evolve from Acoustic- level to Semantic-level in the model. Work performed during an internship at Tencent AI Lab. † Corresponding Author

    work page

  4. [4]

    However, re- search in this area has been hindered by the limitations in data access [6]

    Related work Music Information Retrieval (MIR) tasks primarily aims to extract useful information from raw audio signals, including Genre classification, Key detection, and so on. However, re- search in this area has been hindered by the limitations in data access [6]. To address this challenge, self-supervised meth- ods [4] have been introduced to the MI...

    work page

  5. [5]

    Layer-wise analysis provides us with a detailed perspective on the relationship between the performance of each specific layer and each specific task

    Conclusion We evaluate music SSL models, including MusicFM and MuQ, across 14 tasks, showing their comprehensive advantages. Layer-wise analysis provides us with a detailed perspective on the relationship between the performance of each specific layer and each specific task. It also reveals a transition from acoustic to semantic features. Future work will...

    work page

  6. [6]

    Foundation Model The foundation models are pre-trained on a large amount of un- labeled data to learn high-dimensional representations of audio

    Analysis Methods and Details 3.1. Foundation Model The foundation models are pre-trained on a large amount of un- labeled data to learn high-dimensional representations of audio. It aims to compress and denoise long raw audio signals, retain- ing as much information in signals as possible. The training objective is to predict tokens that have been randoml...

    arXiv 2025

  7. [7]

    SSL outperforms low-level features

    Findings 4.1. SSL outperforms low-level features. According to Table 1, MusicFM model and MuQ model out- perform base features on all of the tasks. In most tasks, the performance of basic features is significantly inferior to that of SSL models. Notably, in the structural analysis task, the basic features achieved comparable performance, likely because th...

    work page

  8. [8]

    wav2vec 2.0: A framework for self-supervised learning of speech representations

    Alexei Baevski, Yuhao Zhou, Abdelrahman Mohamed, and Michael Auli. wav2vec 2.0: A framework for self-supervised learning of speech representations. Advances in neural informa- tion processing systems, 33:12449–12460, 2020

    work page 2020

Show all 40 references
  1. [9]

    Mert: Acoustic music un- derstanding model with large-scale self-supervised training, 2023

    Yizhi Li, Ruibin Yuan, Ge Zhang, et al. Mert: Acoustic music un- derstanding model with large-scale self-supervised training, 2023

    2023

  2. [10]

    base on BEST-RQ [11], proving its generalization ability on sequence-level and frame-level MIR tasks. Some research has been dedicated to analyzing the reasons behind the high performance of speech SSL models [12], and there are studies that have conducted layer-wise analysis ...

  3. [11]

    A foundation model for music informatics

    Minz Won, Yun-Ning Hung, and Duc Le. A foundation model for music informatics. arXiv preprint arXiv:2311.03318, 2023

    2023 arXiv

  4. [12]

    Contrastive learn- ing of musical representations

    Janne Spijkervet and JohnAshley Burgoyne. Contrastive learn- ing of musical representations. Cornell University - arXiv,Cornell University - arXiv, Mar 2021

    2021

  5. [13]

    Codified audio language modeling learns useful representations for music information retrieval

    Rodrigo Castellon, Chris Donahue, and Percy Liang. Codified audio language modeling learns useful representations for music information retrieval. Cornell University - arXiv,Cornell Univer- sity - arXiv, Jul 2021

    2021

  6. [14]

    Supervised and unsuper- vised learning of audio representations for music understanding

    Matthew C McCallum, Filip Korzeniowski, Sergio Oramas, Fa- bien Gouyon, and Andreas F Ehmann. Supervised and unsuper- vised learning of audio representations for music understanding. arXiv preprint arXiv:2210.03799, 2022

    2022 arXiv

  7. [15]

    Freeman, Jessie Wang, Sherry Cai, and KatherineM

    Wenqin Chen, Jessica Keast, Jordan Moody, Corinne Moriarty, Felicia Villalobos, Virtue Winter, Xueqi Zhang, Xuanqi Lyu, E.C. Freeman, Jessie Wang, Sherry Cai, and KatherineM. Kin- naird. Data usage in mir: History & future recommendations. International Symposium/Conference on...

    2019

  8. [16]

    wav2vec: Unsupervised pre-training for speech recognition

    Steffen Schneider, Alexei Baevski, Ronan Collobert, and Michael Auli. wav2vec: Unsupervised pre-training for speech recognition. In Interspeech 2019, Sep 2019

    2019

  9. [17]

    Hubert: Self-supervised speech representation learning by masked prediction of hidden units

    Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, and Abdelrahman Mohamed. Hubert: Self-supervised speech representation learning by masked prediction of hidden units. IEEE/ACM Transactions on Audio, Speech, and Language Processing , pa...

    2021

  10. [18]

    Muq: Self- supervised music representation learning with mel residual vector quantization

    Haina Zhu, Yizhi Zhou, Hangting Chen, Jianwei Yu, Ziyang Ma, Rongzhi Gu, Yi Luo, Wei Tan, and Xie Chen. Muq: Self- supervised music representation learning with mel residual vector quantization. arXiv preprint arXiv:2501.01108, 2025

    2025 arXiv

  11. [19]

    Self-supervised learning with random-projection quantizer for speech recognition

    Chung-Cheng Chiu, James Qin, Yu Zhang, Jiahui Yu, and Yonghui Wu. Self-supervised learning with random-projection quantizer for speech recognition. In ICML, pages 3915–3924, 2022

    2022

  12. [20]

    Why does self-supervised learning for speech recognition benefit speaker recognition? arXiv preprint arXiv:2204.12765, 2022

    Sanyuan Chen, Yu Wu, Chengyi Wang, Shujie Liu, Zhuo Chen, Peidong Wang, Gang Liu, Jinyu Li, Jian Wu, Xiangzhan Yu, et al. Why does self-supervised learning for speech recognition benefit speaker recognition? arXiv preprint arXiv:2204.12765, 2022

    2022 arXiv

  13. [21]

    Layer-wise analysis of a self-supervised speech representation model

    Ankita Pasad, Ju-Chieh Chou, and Karen Livescu. Layer-wise analysis of a self-supervised speech representation model. In2021 IEEE Automatic Speech Recognition and Understanding Work- shop (ASRU), Dec 2021

    2021

  14. [22]

    Liu, Cheng- I Lai, Haibin Wu, Jiatong Shi, Xuankai Chang, Hsiang-Sheng Tsai, Wen-Chin Huang, Tzu hsun Feng, Po-Han Chi, Yist Y

    Shu wen Yang, Heng-Jui Chang, Zili Huang, Andy T. Liu, Cheng- I Lai, Haibin Wu, Jiatong Shi, Xuankai Chang, Hsiang-Sheng Tsai, Wen-Chin Huang, Tzu hsun Feng, Po-Han Chi, Yist Y . Lin, Yung-Sung Chuang, Tzu-Hsien Huang, Wei-Cheng Tseng, Kushal Lakhotia, Shang-Wen Li, Abdelrahma...

    2024

  15. [23]

    Marble: Music audio representation benchmark for universal evaluation

    Ruibin Yuan, Yinghao Ma, Yizhi Li, et al. Marble: Music audio representation benchmark for universal evaluation. arXiv preprint arXiv:2306.10548, 2023

    2023 arXiv

  16. [24]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Confer- ence of the North , Jan 2019

    2019

  17. [25]

    Advances in residual vector quantization: A review

    Christopher F Barnes, Syed A Rizvi, and Nasser M Nasrabadi. Advances in residual vector quantization: A review. IEEE trans- actions on image processing, 5(2):226–262, 1996

    1996

  18. [26]

    The million song dataset

    Thierry Bertin-Mahieux, Daniel PW Ellis, Brian Whitman, and Paul Lamere. The million song dataset. In ISMIR, 2011

    2011

  19. [27]

    Relations between two sets of variates

    Harold Hotelling. Relations between two sets of variates. In Breakthroughs in statistics: methodology and distribution , pages 162–190. Springer, 1992

    1992

  20. [28]

    Svcca: Singular vector canonical correlation analysis for deep learning dynamics and interpretability

    Maithra Raghu, Justin Gilmer, Jason Yosinski, and Jascha Sohl- Dickstein. Svcca: Singular vector canonical correlation analysis for deep learning dynamics and interpretability. Cornell Univer- sity - arXiv,Cornell University - arXiv , Jun 2017

    2017

  21. [29]

    A neural network that finds a naturalistic so- lution for the production of muscle activity.Nature Neuroscience, page 1025–1033, Jul 2015

    David Sussillo, Mark M Churchland, Matthew T Kaufman, and Krishna V Shenoy. A neural network that finds a naturalistic so- lution for the production of muscle activity.Nature Neuroscience, page 1025–1033, Jul 2015

    2015

  22. [30]

    Morcos, Maithra Raghu, and Samy Bengio

    AriS. Morcos, Maithra Raghu, and Samy Bengio. Insights on rep- resentational similarity in neural networks with canonical correla- tion. Neural Information Processing Systems,Neural Information Processing Systems, Jun 2018

    2018

  23. [31]

    Neu- ral audio synthesis of musical notes with wavenet autoencoders

    Jesse Engel, Cinjon Resnick, Adam Roberts, Sander Dieleman, Mohammad Norouzi, Douglas Eck, and Karen Simonyan. Neu- ral audio synthesis of musical notes with wavenet autoencoders. In International Conference on Machine Learning , pages 1068–

  24. [32]

    V ocalset: A singing voice dataset

    Julia Wilkins, Prem Seetharaman, Alison Wahl, and Bryan Pardo. V ocalset: A singing voice dataset. In ISMIR, pages 468–474, 2018

    2018

  25. [33]

    Tzanetakis and P

    G. Tzanetakis and P. Cook. Musical genre classification of audio signals. IEEE Transactions on Speech and Audio Processing, page 293–302, Jul 2002

    2002

  26. [34]

    Caro, Erik M

    Mohammad Soleymani, Micheal N. Caro, Erik M. Schmidt, Cheng-Ya Sha, and Yi-Hsuan Yang. 1000 songs for emotional analysis of music. In Proceedings of the 2nd ACM international workshop on Crowdsourcing for multimedia, Oct 2013

    2013

  27. [35]

    The harmonix set: Beats, downbeats, and functional segment annotations of western popular music

    Oriol Nieto, Matthew C McCallum, Matthew EP Davies, Andrew Robertson, Adam M Stark, and Eran Egozy. The harmonix set: Beats, downbeats, and functional segment annotations of western popular music. In ISMIR, pages 565–572, 2019

    2019

  28. [36]

    Evaluation of algorithms using games: The case of mu- sic tagging

    Edith Law, Kris West, Michael I Mandel, Mert Bay, and J Stephen Downie. Evaluation of algorithms using games: The case of mu- sic tagging. In ISMIR, pages 387–392. International Society for Music Information Retrieval (ISMIR), 2009

    2009

  29. [37]

    Two data sets for tempo estimation and key detection in electronic dance music annotated from user corrections

    Peter Knees, ´Angel Faraldo P ´erez, Herrera Boyer, Richard V ogl, Sebastian B¨ock, Florian H¨orschl¨ager, Mickael Le Goff, et al. Two data sets for tempo estimation and key detection in electronic dance music annotated from user corrections. In Proceedings of the 16th Interna...

    2015

  30. [38]

    Mir eval: A transparent implementation of common mir metrics

    Colin Raffel, Brian McFee, Eric J Humphrey, Justin Salamon, Oriol Nieto, Dawen Liang, Daniel PW Ellis, and C Colin Raffel. Mir eval: A transparent implementation of common mir metrics. In ISMIR, volume 10, page 2014, 2014

    2014

  31. [39]

    Deep contextualized word representations

    Matthew Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. Deep contextualized word representations. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human L...

    2018

  32. [40]

    Comparative layer- wise analysis of self-supervised speech models

    Ankita Pasad, Bowen Shi, and Karen Livescu. Comparative layer- wise analysis of self-supervised speech models. In ICASSP 2023- 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5. IEEE, 2023

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.