Pith. sign in

REVIEW 4 major objections 4 minor 44 references

MuScriptor: An Open Model for Multi-Instrument Music Transcription

T0 review · 4 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

Pith's one-line read This paper claims that a decoder-only transformer trained on real music audio with aligned note annotations, after synthetic pre-training and reinforcement-learning post-training, can transcribe multi-instrument recordings across genres wit

desk verdict Solid open-weight transcription model with real data at scale; the headline 48.2 Multi-F1 is provisional because the test set shares the alignment pipeline with the training data. read the letter →

arxiv 2607.08168 v2 pith:DJJU7PHE submitted 2026-07-09 cs.SD cs.LG

classification cs.SDcs.LG
keywords automaticmusictranscriptionmulti-instrumentopen-weightmodelsyntheticdatapre-trainingreinforcementlearningpost-traininginstrumentconditioningdecoder-onlytransformer
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

The authors set out to show that synthetic data alone is not enough for general-purpose music transcription, but a training recipe of large-scale synthetic pre-training, fine-tuning on 11,000 hours of real recordings, and a policy-gradient post-training step produces a model that far exceeds prior baselines. Each stage contributes: synthetic data helps when real data is scarce, real data drives the largest gains, and reinforcement learning on a small high-quality set adds further improvement. The paper also introduces instrument-presence conditioning, which lets users customize transcriptions and stabilizes segment predictions. If the claims hold, the music information retrieval community gets both a usable open-weight transcriber and a clear data-and-training recipe for building one.

What carries the argument

The key machinery is the training recipe rather than a new architecture: a plain decoder-only transformer over mel-spectrogram input and MIDI-like event tokens, pre-trained on synthetic audio rendered from 1.45M MIDI files, fine-tuned on 170k real recordings aligned to note annotations via audio-score synchronization, and post-trained with a reinforcement-learning objective that rewards note-level F1. A learned instrument-conditioning prefix lets the user specify which instruments to transcribe and stabilizes segment predictions.

What would settle it

Have independent musicians hand-transcribe a random sample of test tracks and compare MuScriptor's output against those human labels; if the multi-instrument F1 falls materially below the reported 48.2, the claimed gains are inflated by alignment bias in the reference annotations.

Watch

Extended reading notes

Core claim

MuScriptor is a 1.3B-parameter decoder-only transformer that maps a mel-spectrogram to a MIDI-like token sequence, optionally conditioned on which instruments are present. Trained on 1.45 million synthetic MIDI files, then 170,000 real recordings with synchronized note annotations, then refined with a group-relative policy-gradient objective on 300 manually verified high-quality pieces, it reaches a multi-instrument F1 of 48.2, onset F1 of 54.4, and frame F1 of 73.3 on the authors' test set—more than doubling the multi-instrument F1 of the strongest prior baseline. The paper's central claim is that this combination of synthetic pre-training, real-data fine-tuning, and reinforcement-learning

Load-bearing premise

The load-bearing premise is that the audio-to-note alignment used to build both training and test data is note-accurate across all genres; if alignment errors correlate with musical style, the measured gains partly reflect noise in the reference labels.

Editorial extensions

If this is right

  • If the recipe transfers, other music information retrieval tasks such as chord, key, and melody extraction could adopt the same data-scale strategy instead of relying on architectural tweaks.
  • The 11,000-hour real aligned dataset becomes a reusable public resource that lowers the barrier for future transcription research.
  • Instrument-conditioned transcription enables user-facing tools that extract specific instruments, such as bass or vocal lines, from a full mix.
  • Reinforcement-learning post-training opens a new axis for improving transcription beyond standard supervised teacher forcing.

Reading between the lines

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

  • The reported numbers rest on a test set built from the same audio-score synchronization pipeline used for training; a fully independent, human-transcribed evaluation set could yield lower estimates, especially for onset and offset precision, noting that the Bach10 onset score actually dropped from 59.8 to 43.1 relative to the baseline, a hint of style-dependent alignment issues.
  • If the synchronization quality scales with data, further improvements in alignment—or larger real aligned corpora—could push multi-instrument transcription well beyond the plateau reported here.
  • The instrument-conditioning mechanism could be used interactively, letting a user toggle instruments on and off to separate a track by ear, effectively turning the model into a source-separation-style tool despite not being explicitly trained for separation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper presents MuScriptor, a decoder-only transformer for multi-instrument music transcription that is trained on 1.45M synthetic MIDI files and 170k real recordings (11k hours) with audio-symbolic alignments, then post-trained with a GRPO-style reinforcement learning objective using 300 manually verified tracks. The model optionally conditions on instrument presence, and the authors release model weights and inference code. The central empirical claim is that MuScriptor is the first effective open-weight multi-instrument transcription model, supported by large gains over YourMT3+ on a curated in-house test set D_Test (e.g., Multi F1 48.2 vs 21.9) and by improvements on several external benchmarks.

Significance. If the central claim holds, this is a significant contribution: the open-weight model, the data recipe combining synthetic pre-training with large-scale real fine-tuning, the RL post-training, and the instrument conditioning are all useful to the AMT community. The internal ablations are consistent, the synthetic-data analysis is informative, and the model release is a concrete asset. However, the main evaluation depends on automatically synchronized annotations that are shared between training and test, and several independent benchmarks show regressions relative to the baseline. The paper's value is therefore conditional on validating the alignment pipeline and on presenting the D_Test results as in-domain rather than as evidence of general-purpose transcription quality.

major comments (4)
  1. [§3.1, Table 1] The headline comparison is evaluated on D_Test, which is selected from D_Real, whose annotations are produced by audio-symbolic synchronization (linear bar-line interpolation or DTW with chroma/onset features). The stated filters only constrain global warping distance and a maximum time-dilation factor; they do not validate note-level accuracy. If synchronization errors are systematic, both training targets and D_Test ground truth are corrupted in a correlated way, so the reported 48.2 vs 21.9 Multi F1 gap can overstate real-world performance. The independent benchmarks in Table 2 are mixed: Bach10 onset F1 drops from 59.8 (YourMT3+) to 43.1, and RWC-C/RWC-R also show regressions. Please provide a quantitative validation of the alignment pipeline (e.g., note-level agreement on a manually annotated sample), state whether D_Test was manually verified, and report per-genre and per-dataset i
  2. [§3.1 (D_RL/D_Test disjointness)] It is not stated whether D_RL and D_Test are disjoint. D_RL is described as 300 tracks selected from D_Real, and D_Test as 372 tracks subsequently selected from D_Real; the text says only that D_Test tracks are removed from D_Real and that similar titles are removed from training sets. If D_RL tracks were not excluded from the D_Test selection, the RL post-training set could overlap with the test set, inflating the D_RL row of Table 1. Please state explicitly that D_RL and D_Test are disjoint, and describe how overlap was checked beyond title matching.
  3. [Tables 1 and 3] The headline MuScriptor results use ground-truth instrument conditioning (the actual instruments present in the track), while the YourMT3+ baseline is not given this oracle information. Table 3 shows conditioning contributes about 1.8 Multi F1 on a D_Real-only model, so it does not explain most of the Table 1 gap, but it is still a confound in the claim of being the first effective open-weight model. Please report the main comparison with MuScriptor evaluated without oracle conditioning (using conditioning dropout or predicted instruments), and, if possible, include an instrument-conditioned baseline.
  4. [§3.3, §4.2.3] The primary D_Test metrics are computed after deleting overlapping same-pitch same-instrument notes that the tokenizer cannot represent. The authors disclose this and report that keeping these notes lowers Multi F1 from 48.2 to 42.0 (α_CFG=1). Since such overlaps are common in real mixes, the headline number is not a direct measure of real-world transcription quality. The conclusion that MuScriptor is 'effective' for general-purpose transcription should be qualified, and the central claims should be presented on both the filtered and unfiltered test sets, not only on the filtered one.
minor comments (4)
  1. [Table 1 and Figure 4] No error bars or multiple runs are reported for the central numbers. Given the D_Test size and the stochasticity of RL post-training, bootstrap confidence intervals or results across seeds would materially strengthen the comparison.
  2. [Eq. (3)] The notation L_CE(θ, τ; y) uses l_θ(y_t), which is called a logit, but the expression is a temperature-scaled cross-entropy term. Clarify that this is a token-level loss and how it relates to the REINFORCE estimator's log-probability.
  3. [§4.1.2] The sentence attributing lower onset/offset scores to 'difficulty of annotating precise note on- and offsets for certain styles' does not explain the Bach10 onset regression. A specific analysis of the Bach10 case would be informative.
  4. [§3.1 / Data release] The paper is titled 'Open Model' and releases weights and inference code, but D_Real and D_Test are internal/commercial and not released. Please document licenses and consider releasing a small public evaluation subset to make the benchmark reproducible.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: all headline results are direct measurements on held-out data; minor self-citations are not load-bearing.

full rationale

MuScriptor's central claims are empirical training/evaluation results, not a derivation chain. The model is trained with cross-entropy on D_Synth/D_Real and a REINFORCE-style objective on D_RL, then evaluated by mir_eval metrics on a held-out D_Test and on external benchmarks. No fitted parameter is renamed as a prediction: the RL reward is a training objective, not a test result, and no claimed result is defined in terms of the quantity it purports to predict. The use of D_Test drawn from the same audio-symbolic synchronization pipeline as D_Real is an annotation-quality/independence concern, but it is not circular, because the reported F1 scores are computed directly from model outputs versus reference annotations and the paper also reports independent external benchmarks (Table 2), including a candid regression on Bach10 onset (43.1 vs 59.8) that it does not hide. Self-citations to [20] (Sync Toolbox, co-authored by M. Krause), [22] (MusicGen architecture, co-authored by A. Défossez), and [31] (Encodec) are tool/method citations for standard components; none is invoked as a uniqueness theorem, and none smuggles in the conclusion via an unverified ansatz. The paper also self-flags metric limitations in §4.2.3, where dropping unrepresentable overlapping notes lowers onset F1 from 60.4 to 51.8 and notes that 'A different tokenization scheme is needed.' Therefore no circular step meeting the evidentiary bar is present; at most there are minor self-citations that are not load-bearing.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper is an engineering contribution; it introduces no new physical or mathematical entities. Its load-bearing premises are the accuracy of its data-curation pipeline, the representational sufficiency of the tokenization, and the representativeness of its self-curated test set; its free choices are training and inference hyperparameters plus curation thresholds.

free parameters (4)
  • Classifier-free guidance strength α_CFG = 2
    Chosen as it improved most configurations in Table 1; a hand-picked inference hyperparameter that affects scores by several F1 points.
  • GRPO sampling temperature τ and group size G = τ=0.75, G=8
    Set for RL post-training without reported sweep; affects reward signal and training stability.
  • Audio-symbolic alignment filters = not specified
    Threshold on DTW warping distance and 'maximum time dilation factor (after 8 s, must progress 1 s)' define which 170k tracks enter D_Real and hence determine training targets and evaluation data.
  • Training schedule hyperparameters = lr 1e-4, 1M steps, batch 64
    Standard choices; affect final quality but not load-bearing.
assumptions (5)
  • domain assumption The audio-symbolic synchronization pipeline (Sync Toolbox DTW with chroma and onset features) produces note-accurate alignments for 170k real tracks.
    Invoked in §3.1 to build D_Real; if alignments are inaccurate or biased, both training supervision and D_Test ground truth are compromised.
  • domain assumption MT3 tokenization with the 36-instrument MT3_FULL_PLUS taxonomy can represent all notes in the data; overlapping same-pitch same-instrument notes are rare enough to remove from evaluation.
    Invoked in §3.2–3.3; the model is structurally unable to emit simultaneous same-pitch same-instrument notes, and the eval removes them.
  • domain assumption Manually verified D_Test (372 tracks) and D_RL (300 tracks) annotations are accurate ground truth and representative of hard multi-instrument music.
    Invoked in §3.1; the headline comparison rests on this self-curated set.
  • standard math GRPO/REINFORCE with group-relative advantage is a valid training objective for this sequence model.
    Invoked in §3.4; standard policy-gradient machinery, no formal guarantee of improvement.
  • domain assumption Mel-spectrogram with 512 mel bins at 100 Hz frame rate retains enough information for transcription.
    Invoked in §3.2; empirically validated against CQT/Encodec/MERT in Table 5.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MuScriptor: An Open Model for Multi-Instrument Music Transcription." pith.science (2026). https://pith.science/paper/DJJU7PHE

@misc{pith2026260708168,
  author       = {Pith},
  title        = {Pith review of: MuScriptor: An Open Model for Multi-Instrument Music Transcription},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DJJU7PHE}},
  note         = {Machine review of arXiv:2607.08168}
}
read the original abstract

Existing methods for automatic music transcription are often limited to single-instrument recordings or fail on complex, real music mixes. Although previous work utilizes synthetic training data, the resulting models generalize poorly, leading to largely unusable transcription output in realistic, multi-instrument settings. In this work, we analyze the effectiveness of synthetic data for pre-training while combining it with fine-tuning on real music audio and post-training using reinforcement learning. We further introduce conditioning on instrument presence to customize transcriptions. Finally, we release MuScriptor, an open-weight multi-instrument music transcription model that works on real-world music recordings from across a diverse range of musical genres.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 3 linked inside Pith

  1. [1]

    MuScriptor: An Open Model for Multi-Instrument Music Transcription

    INTRODUCTION The task of Automatic Music Transcription (AMT) con- sists of converting an audio recording of a piece of music into some kind of symbolic representation, typically MIDI. While significant progress has been made in transcrib- ing single-instrument recordings (specifically for piano [1], guitar [2], and drums [3, 4]), general-purpose transcrip...

  2. [2]

    To train it, we gather a dataset of 170k recordings (11k hours) with aligned note annotations that covers most musical genres, alongside a synthetic dataset of 1.45M MIDIs

    An Open-Weight Transcription Model:We re- lease MuScriptor, a model (including weights and infer- ence code) 1 for general-purpose multi-instrument music transcription across diverse musical styles. To train it, we gather a dataset of 170k recordings (11k hours) with aligned note annotations that covers most musical genres, alongside a synthetic dataset o...

  3. [3]

    Analysis of Synthetic Pre-training:We provide a comparative study on the benefits and limitations of pre- training on synthetic data versus training on real-world recordings with aligned note annotations

  4. [4]

    †Equal contribution 1 https://github.com/muscriptor/muscriptor arXiv:2607.08168v2 [cs.SD] 3 Aug 2026 Figure 1

    Alignment via Reinforcement Learning:We apply a GRPO-like algorithm to our model to align it with a small dataset of high-quality transcriptions. †Equal contribution 1 https://github.com/muscriptor/muscriptor arXiv:2607.08168v2 [cs.SD] 3 Aug 2026 Figure 1. Piano roll for the guitar notes in a song from our test setD Test. Blue indicates true positives (a ...

  5. [5]

    Instrument-conditioned Transcription:By option- ally conditioning on instrument presence, we stabilize pre- dictions across different segments of a music recording and allow users to customize their transcription output

  6. [6]

    RELA TED WORK AMT models historically evolved from highly specialized instrument specific architectures towards generalized se- quence modeling paradigms. Early research focused on single-instrument scenarios (most notably solo piano) us- ing traditional methods such as hidden markov models [6], non negative matrix factorization [7], or support vector mac...

  7. [7]

    Extend- ing this framework, we investigate training on both large- scale synthetic and real music audio datasets (Section 3.1)

    METHOD Similar to previous work, our model performs transcription by autoregressively predicting a MIDI-like token sequence given a mel-spectrogram of a short audio segment. Extend- ing this framework, we investigate training on both large- scale synthetic and real music audio datasets (Section 3.1). We further utilize reinforcement learning to improve re...

  8. [8]

    Unless stated otherwise, all results are for the 1.3B parameters model with instrument conditioning

    EXPERIMENTAL RESULTS In our experiments, we generally pre-train our model on DSynth, fine-tune onD Real, post-train onD RL, and evaluate onD Test. Unless stated otherwise, all results are for the 1.3B parameters model with instrument conditioning. 4.1 Main Results Our main results are summarized in Table 1. We observe that each training stage improves res...

Show all 44 references
  1. [9]

    To train MuScriptor, we collected synthetic and real music audio datasets with aligned note annota- tions, and used a reinforcement learning approach for post- training refinement

    CONCLUSION In this work, we presented MuScriptor, an open weight model for general-purpose, multi-instrument music tran- scription. To train MuScriptor, we collected synthetic and real music audio datasets with aligned note annota- tions, and used a reinforcement learning appr...

  2. [10]

    Enabling factorized piano music modeling and generation with the MAESTRO dataset,

    C. Hawthorne, A. Stasyuk, A. Roberts, I. Simon, C. A. Huang, S. Dieleman, E. Elsen, J. H. Engel, and D. Eck, “Enabling factorized piano music modeling and generation with the MAESTRO dataset,” inICLR, 2019. [Online]. Available: https: //openreview.net/forum?id=r1lYRjC9F7

  3. [11]

    GuitarSet: A dataset for guitar transcription,

    Q. Xi, R. M. Bittner, J. Pauwels, X. Ye, and J. P. Bello, “GuitarSet: A dataset for guitar transcription,” inISMIR, 2018. [Online]. Available: http://ismir2018.ircam.fr/doc/pdfs/188_Paper.pdf

  4. [12]

    Deep unsupervised drum transcription,

    K. Choi and K. Cho, “Deep unsupervised drum transcription,” inISMIR, 2019. [Online]. Available: http://archives.ismir.net/ismir2019/paper/000020.pdf

  5. [13]

    In- creasing drum transcription vocabulary us- ing data synthesis,

    M. Cartwright and J. P. Bello, “In- creasing drum transcription vocabulary us- ing data synthesis,” inDAFx, 2018. [On- line]. Available: https://markcartwright.com/files/ cartwright2018increasing.pdf

  6. [14]

    MT3: multi-task multitrack music transcription,

    J. Gardner, I. Simon, E. Manilow, C. Hawthorne, and J. Engel, “MT3: multi-task multitrack music transcription,” inICLR, 2022. [Online]. Available: https://openreview.net/forum?id=iMSjopcOn0p

  7. [15]

    Automatic transcription of piano music,

    C. Raphael, “Automatic transcription of piano music,” inISMIR, 2002. [Online]. Available: http://ismir2002.ismir.net/proceedings/02-FP01-2.pdf

  8. [16]

    Non-negative matrix factorization for polyphonic music transcription,

    P. Smaragdis and J. C. Brown, “Non-negative matrix factorization for polyphonic music transcription,” in WASPAA, 2003, pp. 177–180. [Online]. Available: https://ieeexplore.ieee.org/document/1285860

  9. [17]

    A discriminative model for polyphonic piano transcription,

    G. E. Poliner and D. P. W. Ellis, “A discriminative model for polyphonic piano transcription,”EURASIP J. Adv. Signal Process., 2007. [Online]. Available: https://doi.org/10.1155/2007/48317

  10. [18]

    Polyphonic piano note transcription with recurrent neural networks,

    S. Böck and M. Schedl, “Polyphonic piano note transcription with recurrent neural networks,” in ICASSP, 2012, pp. 121–124. [Online]. Available: https://doi.org/10.1109/ICASSP.2012.6287832

  11. [19]

    An end- to-end neural network for polyphonic piano music transcription,

    S. Sigtia, E. Benetos, and S. Dixon, “An end- to-end neural network for polyphonic piano music transcription,”IEEE ACM Trans. Audio Speech Lang. Process., vol. 24, no. 5, pp. 927–939,

  12. [20]

    Sync toolbox: A python package for efficient, robust, and accurate music synchronization,

    M. Müller, Y . Özer, M. Krause, T. Prätzlich, and J. Driedger, “Sync toolbox: A python package for efficient, robust, and accurate music synchronization,” Journal of Open Source Software, vol. 6, no. 64, p. 3434, 2021. [Online]. Available: https://doi.org/ 10.21105/joss.03434

  13. [21]

    Onsets and frames: Dual-objective piano transcription,

    C. Hawthorne, E. Elsen, J. Song, A. Roberts, I. Simon, C. Raffel, J. H. Engel, S. Oore, and D. Eck, “Onsets and frames: Dual-objective piano transcription,” inISMIR, 2018. [Online]. Available: http://ismir2018.ircam.fr/doc/pdfs/19_Paper.pdf

  14. [22]

    Sequence-to-sequence piano transcrip- tion with transformers,

    C. Hawthorne, I. Simon, R. Swavely, E. Manilow, and J. Engel, “Sequence-to-sequence piano transcrip- tion with transformers,” inISMIR, 2021. [Online]. Available: https://archives.ismir.net/ismir2021/paper/ 000030.pdf

  15. [23]

    YourMT3+: multi-instrument mu- sic transcription with enhanced transformer ar- chitectures and cross-dataset stem augmentation,

    S. Chang, E. Benetos, H. Kirchhoff, and S. Dixon, “YourMT3+: multi-instrument mu- sic transcription with enhanced transformer ar- chitectures and cross-dataset stem augmentation,” inMLSP, 2024. [Online]. Available: https: //doi.org/10.1109/MLSP58920.2024.10734819

  16. [24]

    Advancing multi-instrument music tran- scription: Results from the 2025 AMT challenge,

    O. Chaturvedi, K. Bhardwaj, T. Gondil, B. S.- H. Chou, K. Y .-J. Yun, Y .-H. Lu, Y . Yan, and S. Chang, “Advancing multi-instrument music tran- scription: Results from the 2025 AMT challenge,” inNeurIPS AI4Music Workshop, 2025. [Online]. Available: https://arxiv.org/abs/2603.27528

  17. [25]

    A foundation model for music informatics,

    M. Won, Y . Hung, and D. Le, “A foundation model for music informatics,” inICASSP, 2024, pp. 1226–

  18. [26]

    DeepSeekMath: pushing the limits of mathematical reasoning in open language models,

    Z. Shao, P. Wang, Q. Zhu, R. Xu, J. Song, X. Bi, H. Zhang, M. Zhang, Y . K. Li, Y . Wu, and D. Guo, “DeepSeekMath: pushing the limits of mathematical reasoning in open language models,”arXiv, 2024. [Online]. Available: https://arxiv.org/abs/2402.03300

  19. [27]

    Unaligned supervi- sion for automatic music transcription in the wild,

    B. Maman and A. H. Bermano, “Unaligned supervi- sion for automatic music transcription in the wild,” in ICML, 2022, pp. 14 918–14 934. [Online]. Available: https://proceedings.mlr.press/v162/maman22a.html

  20. [28]

    To- wards realistic synthetic data for automatic drum transcription,

    P. Melucci, P. Merialdo, and T. Akama, “To- wards realistic synthetic data for automatic drum transcription,”arXiv, 2026. [Online]. Available: https://arxiv.org/abs/2601.09520

  21. [29]

    Learn from virtual guitar: A comparative analysis of automatic guitar transcription using synthetic and real audio,

    Y . Kusaka and A. Maezawa, “Learn from virtual guitar: A comparative analysis of automatic guitar transcription using synthetic and real audio,” in WASPAA, 2025. [Online]. Available: https://doi.org/ 10.1109/W ASPAA66052.2025.11230954

  22. [30]

    Learning-based methods for com- paring sequences, with applications to Audio-to- MIDI alignment and matching,

    C. Raffel, “Learning-based methods for com- paring sequences, with applications to Audio-to- MIDI alignment and matching,” Ph.D. dissertation, Columbia University, USA, 2016. [Online]. Available: https://doi.org/10.7916/D8N58MHV

  23. [31]

    High fidelity neural audio compression,

    A. Défossez, J. Copet, G. Synnaeve, and Y . Adi, “High fidelity neural audio compression,”Trans. Mach. Learn. Res., 2023. [Online]. Available: https: //openreview.net/forum?id=ivCd8z8zR2

  24. [32]

    Using activation functions for improving measure- level audio synchronization,

    Y . Özer, M. Istvanek, V . Arifi-Müller, and M. Müller, “Using activation functions for improving measure- level audio synchronization,” inISMIR, 2022, pp. 749–

  25. [34]

    Simple and control- lable music generation,

    J. Copet, F. Kreuk, I. Gat, T. Remez, D. Kant, G. Syn- naeve, Y . Adi, and A. Défossez, “Simple and control- lable music generation,” inNeurIPS, 2023. [Online]. Available: http://papers.nips.cc/paper_files/paper/ 2023/hash/94b472a1842cd7c56dcb125fb2765fbd- Abstract-Conference.html

  26. [35]

    Classifier-free diffusion guidance,

    J. Ho and T. Salimans, “Classifier-free diffusion guidance,”arXiv, 2022. [Online]. Available: https: //arxiv.org/abs/2207.12598

  27. [36]

    mir_eval: A transparent implementation of common MIR metrics,

    C. Raffel, B. McFee, E. J. Humphrey, J. Salamon, O. Nieto, D. Liang, and D. P. W. Ellis, “mir_eval: A transparent implementation of common MIR metrics,” inISMIR, 2014, pp. 367–372. [Online]. Available: https://www.ee.columbia.edu/~dpwe/pubs/ RaffMHS14-mireval.pdf

  28. [37]

    Simple statistical gradient-following algorithms for connectionist reinforcement learning,

    R. J. Williams, “Simple statistical gradient-following algorithms for connectionist reinforcement learning,” Mach. Learn., vol. 8, pp. 229–256, 1992. [Online]. Available: https://doi.org/10.1007/BF00992696

  29. [39]

    Multi- ple fundamental frequency estimation by mod- eling spectral peaks and non-peak regions,

    Z. Duan, B. Pardo, and C. Zhang, “Multi- ple fundamental frequency estimation by mod- eling spectral peaks and non-peak regions,” IEEE Trans. Speech Audio Process., vol. 18, no. 8, pp. 2121–2133, 2010. [Online]. Available: https://doi.org/10.1109/TASL.2010.2042119

  30. [40]

    Dagstuhl ChoirSet: A multitrack dataset for MIR research on choral singing,

    S. Rosenzweig, H. Cuesta, C. Weiß, F. Scherbaum, E. Gómez, and M. Müller, “Dagstuhl ChoirSet: A multitrack dataset for MIR research on choral singing,”Trans. Int. Soc. Music. Inf. Retr ., vol. 3, no. 1, pp. 98–110, 2020. [Online]. Available: https://doi.org/10.5334/tismir.48

  31. [41]

    Score-informed source separation for multichannel orchestral recordings,

    M. Miron, J. J. Carabias-Orti, J. J. Bosch, E. Gómez, and J. Janer, “Score-informed source separation for multichannel orchestral recordings,” Journal of Electrical and Computer Engineering, vol. 2016, no. 1, p. 8363507, 2016. [Online]. Avail- able: https://onlinelibrary.wiley...

  32. [42]

    RWC revisited: Towards a community-driven MIR corpus,

    S. Balke, J. Zeitler, V . Arifi-Müller, B. McFee, T. Nakano, M. Goto, and M. Müller, “RWC revisited: Towards a community-driven MIR corpus,”Trans. Int. Soc. Music. Inf. Retr ., vol. 9, no. 1, pp. 21–35, 2026. [Online]. Available: https://doi.org/10.5334/tismir.326

  33. [44]

    MERT: acoustic music understanding model with large-scale self-supervised training,

    Y . Li, R. Yuan, G. Zhang, Y . Ma, X. Chen, H. Yin, C. Xiao, C. Lin, A. Ragni, E. Benetos, N. Gyenge, R. B. Dannenberg, R. Liu, W. Chen, G. Xia, Y . Shi, W. Huang, Z. Wang, Y . Guo, and J. Fu, “MERT: acoustic music understanding model with large-scale self-supervised training,...

  34. [756]

    Available: https://archives.ismir.net/ ismir2022/paper/000090.pdf

    [Online]. Available: https://archives.ismir.net/ ismir2022/paper/000090.pdf

  35. [1230]

    Available: https://doi.org/10.1109/ ICASSP48485.2024.10448314

    [Online]. Available: https://doi.org/10.1109/ ICASSP48485.2024.10448314

  36. [2016]

    Available: https://doi.org/10.1109/ TASLP.2016.2533858

    [Online]. Available: https://doi.org/10.1109/ TASLP.2016.2533858

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.