Improving Audio Classification by Transitioning from Zero- to Few-Shot

James Taylor; Wolfgang Mack

REVIEW 4 major objections 6 minor 38 references

Improving Audio Classification by Transitioning from Zero- to Few-Shot

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Defining sound classes by a few averaged audio examples beats defining them by text prompts.

desk verdict Per-class averaged CLAP embeddings beat zero-shot prompts, but uncontrolled external baselines and biased support selection make the ESC-50 and FSD50K margins unproven. read the letter →

arxiv 2507.20036 v1 pith:IK3XMEP5 submitted 2025-07-26 cs.SD cs.LG

James Taylor , Wolfgang Mack This is my paper · ORCID

classification cs.SDcs.LG

keywords few-shotlearningzero-shotaudioclassificationCLAPembeddingsprototypeaveragingtextmutualinformationfeatureselectionsoundevent

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

Few-shot classification here means replacing the text prompt that a contrastive audio-language model uses to define a class with a prototype built by averaging a few real audio recordings of that class. The paper claims this substitution consistently beats zero-shot classification: with 50 examples per class, averaged CLAP audio embeddings reach 0.716 accuracy on the internal BBL set versus 0.623 for CLAP23 text prompts, 0.970 on ESC-50 versus 0.948, and 0.579 mAP on FSD50K versus 0.485. The practical payoff is that a small handful of labeled recordings removes the need to hunt for the best text description of a sound class. The paper also reports that simple averaging beats linear-discriminant classifiers and that mutual-information feature selection gives little extra.

What carries the argument

The central object is the per-class averaged audio embedding, a prototype $e_c$ computed from a small set $E_c$ of CLAP audio embeddings for class $c$ via $e_c = \sum_{e_i \in E_c} w_i e_i$, with either uniform weights (arithmetic mean) or $\ell^2$-normalized weights. This prototype replaces the text embedding that zero-shot classification compares against; test embeddings are assigned by minimum distance, using cosine similarity (consistent with CLAP's training) rather than mean squared error. The paper also tests LDA classifiers and mutual-information feature selection as alternatives, but these are secondary: simple prototype averaging is the mechanism that carries the claimed gains.

What would settle it

A direct check is to reproduce the ESC-50 and FSD50K zero-shot baselines with the same CLAP-23 checkpoint and the paper's evaluation protocol; if the reproduced zero-shot accuracy reaches 0.970 on ESC-50 or mAP reaches 0.579 on FSD50K, the claimed few-shot advantage collapses.

Watch

Extended reading notes

Core claim

The paper's discovery is that audio-derived class prototypes are a better reference than text-derived embeddings for contrastive audio classification. Starting from a pretrained CLAP audio encoder, each class is represented by the average of its audio embeddings from a small development set, and test clips are classified by cosine similarity to the closest prototype. Across three datasets this few-shot recipe beats the same model family's zero-shot text classification by 2 to 10 percentage points, with larger margins where text descriptions are confusing, for example improving 'car' accuracy on BBL from near zero to 45 percent. LDA underperforms the average, which the paper attributes to the curse of dimensionality in 1024-dimensional embeddings; fine-tuning CLAP still slightly edges out the few-shot method on ESC-50, but at far higher adaptation cost.

Load-bearing premise

The comparison assumes the zero-shot baselines quoted for ESC-50 and FSD50K were measured with the same CLAP checkpoint, prompt wording, and evaluation protocol as the few-shot results, but those baselines come from external papers and only the BBL baseline was re-run, with manual label remapping.

Editorial extensions

If this is right

With as few as 10 audio examples per class, averaged prototypes already beat the CLAP23 zero-shot baseline on all three datasets.
On ESC-50, fewer than five examples per class are enough for the averaged-prototype method to outperform the zero-shot text baseline.
Raising the number of examples per class from 10 to 50 increases accuracy for the averaging method, while LDA-based classifiers need more examples and stay below averaging.
Mutual-information feature selection gives only a small boost to averaging and helps LDA only at large feature budgets, so simple averaging carries the result.
Because only the audio encoder is used at adaptation time, new classes can be added by supplying a few recordings, without retraining or fine-tuning the contrastive model.

Reading between the lines

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

If the gains survive controlled baseline reproduction, collecting a few recordings per class could replace manual prompt selection and label-to-vocabulary remapping in deployed zero-shot audio systems.
The result is contingent on the quality of the underlying audio encoder; with weaker or domain-mismatched embeddings, averaged prototypes will likely gain less, so the recipe is an add-on to representation quality rather than a substitute.
A natural extension is to compare prototype averaging against linear probing or learned nearest-centroid classifiers on the same CLAP embeddings to isolate how much of the gain comes from the prototype construction itself.
The prototype idea could generalize to hierarchical or overlapping-sound settings by averaging at multiple ontology levels or using mixtures of prototypes; the paper only evaluates flat, largely non-overlapping labels.

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.

Desk Editor's Note

read the letter

James Taylor and Wolfgang Mack report a simple, useful result: for CLAP-based audio classification, averaging the embeddings of 10–50 labeled audio examples per class beats zero-shot text-prompt classification on BBL, ESC-50, and FSD50K by 2–10 percentage points. The method is just prototypical few-shot classification applied to CLAP features, and the paper doesn't cite that literature, but the practical message is clear and the experiments are mostly straightforward.

What's genuinely good: The paper is honest about the fragility of text prompts—they manually remapped labels like 'low rev engines' to 'idling tractor' for BBL, which demonstrates the sensitivity they're trying to mitigate. Figure 1 shows how performance scales with support size, with error bars. The BBL comparison is internally controlled. The AVG result is simple enough to reproduce without code.

The soft spots are real but addressable. The zero-shot baselines for ESC-50 and FSD50K are taken from the original papers rather than re-run with the same CLAP checkpoint and prompts. The ESC-50 margin is only 0.970 vs. 0.948, so a slightly different zero-shot protocol could erase the gain. That's a controlled-comparison problem, and the stress-test is right to flag it. For FSD50K, the support samples are deliberately chosen as those with least class overlap, which gives the few-shot arm an advantage that has no zero-shot analogue; the 9-point mAP gain may partially reflect easier support examples. The MI feature-count K is chosen on the evaluation set (Table 1 caption, Figure 3), which inflates the MI rows; the plain AVG result doesn't depend on K, so that's a minor issue. And there's no code release or detailed FSD50K mAP protocol, which limits reproducibility.

These aren't fatal. The central AVG claim is plausible and consistent with the broader few-shot literature; I'd expect a properly controlled re-run to still show gains on BBL, and possibly smaller gains elsewhere. The paper doesn't overclaim—it explicitly notes that fine-tuning still beats few-shot on ESC-50. The novelty is low, but for practitioners who need a cheap upgrade to deployed zero-shot systems, the finding is valuable.

I'd send it to peer review, but the authors should re-run the external baselines (or at minimum add a sensitivity analysis with different prompts and checkpoints), randomize the FSD50K support selection, and release code. The right venue is a workshop or a practice-oriented conference session, not a flagship methods paper.

I'd bring it to the reading group as a data point on how few-shot baselines should be controlled, but I wouldn't cite it in my own work. The authors are thinking honestly; the problems are in the evaluation protocol, not in the logic.

Referee Report

4 major / 6 minor

Summary. The paper proposes replacing text-derived class embeddings with class-averaged audio embeddings in a few-shot setting for audio classification. Using CLAP 2023 audio embeddings, the authors compute per-class reference vectors from a small number of labeled support examples and classify by nearest-vector matching, optionally preceded by linear discriminant analysis or mutual-information feature selection. They evaluate on an internal BBL dataset, ESC-50, and FSD50K, and report that the averaged-embedding method (FS AVG) exceeds zero-shot CLAP23 by roughly 2 to 10 percentage points, while LDA-based variants are less competitive. The paper also studies the effect of the number of support samples and the role of mutual-information dimensionality reduction.

Significance. If the reported margins are reliable, the central observation is practically useful: a parameter-free average of CLAP audio embeddings from a handful of labeled samples can replace text embeddings for closed-set classification, avoiding prompt engineering and fine-tuning costs. The paper's AVG method is simple, deterministic, and easy to reproduce in principle, and the comparison to a fine-tuned CLAP model in Figure 1 is a useful calibration. The paper is also honest in acknowledging the fragility of text-based labels, e.g., the BBL label remappings in the Section 5.1 footnote. However, the quantitative claim that few-shot 'typically outperforms' zero-shot is not yet fully supported because the zero-shot baselines on two of three datasets are imported from external papers rather than recomputed under the same protocol, and because one few-shot variant selects support samples on a basis that has no zero-shot analogue. The practical significance therefore depends on completing the controlled comparison.

major comments (4)

[Table 1] The zero-shot rows for ESC-50 and FSD50K are taken from external papers, as stated in the table caption, while the few-shot rows are computed by the authors with their own CLAP version, prompt choices, and evaluation pipeline. The claimed margins are small on ESC-50 (0.970 vs. 0.948, i.e., 2.2 pp) and substantial on FSD50K (0.579 vs. 0.485, i.e., 9.4 pp), but they cannot be interpreted as controlled comparisons unless the zero-shot baseline is rerun with the identical CLAP checkpoint and text prompts. The paper's own note that BBL labels had to be manually remapped (e.g., 'low rev engines' to 'idling tractor') demonstrates how sensitive zero-shot scores are to text-side choices. Please rerun CLAP23 zero-shot on ESC-50 and FSD50K under the same protocol, or explicitly justify why the externally reported numbers are directly comparable.
[Section 4.2] For FSD50K, the few-shot support samples are 'specifically choose[n] those with the least class overlap with other files'. This selection rule has no analogue in the zero-shot arm and can make the few-shot task easier by removing ambiguous recordings. Since FSD50K is multi-label and the reported gain is 9.4 pp, the result may reflect support-set selection rather than the few-shot principle. Please report results with random support selection, or justify that the overlap-based selection does not systematically inflate the FSD50K numbers.
[Table 1 caption and Figure 3] The caption states that for MI-based methods 'we use the best K for a given method/dataset pair (see Figure 3)'. Choosing K on the evaluation set is a form of test-set selection, and it overstates the MI+AVG and MI+LDA results. This does not affect the plain AVG rows, which are the central claim, but the MI rows are reported as results rather than as an oracle upper bound. Please use a fixed K or select K on a held-out split, and report the corresponding performance for all K values.
[Section 5.1] The BBL zero-shot baseline, although rerun by the authors, required manual adjustments to align class names with CLAP's text vocabulary. The paper does not report the exact prompt template or the full list of label mappings used for BBL, ESC-50, or FSD50K. Since the central claim is that audio-based few-shot representations are more robust than text-based ones, the exact text protocol is a load-bearing part of the comparison. Please provide the prompt templates, the label mappings, and the specific CLAP 2023 checkpoint identifier used in the experiments.

minor comments (6)

[Abstract and Section 5.1] The abstract says few-shot classification 'typically outperforms' the zero-shot baseline, but LDA with |Ec|=10 underperforms zero-shot in Table 1; please qualify the claim to the averaged-embedding method or to the evaluated regimes.
[Equation (2)] The notation for the class-conditional embedding set is inconsistent: the text uses 'E c' while the equation uses 'Ec'. Please standardize the notation.
[Section 4.2] The mutual-information feature-selection procedure is underspecified: no MI estimator, binning strategy, or tie-handling rule is given, which makes the MI+AVG and MI+LDA results hard to reproduce.
[Figure 2] The confusion-matrix comparison in Figure 2 does not state which support-set size |Ec| is used for the FS AVG result; please add this information to the caption.
[Figure 1] The scaling behaviour as a function of |Ec| is only shown for ESC-50; showing the same curve for BBL or FSD50K would make the few-shot data-efficiency claim more general.
[Reference [26]] Reference [26] is cited for the curse of dimensionality but appears to be a book on dynamic programming; please replace it with an appropriate citation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: few-shot class prototypes are computed directly from audio embeddings and evaluated against external zero-shot baselines; no claim reduces to its own inputs.

full rationale

The paper is an empirical comparison rather than a derivation. Equation (2) defines few-shot class prototypes as weighted averages of audio embeddings, and Equation (3) assigns a test embedding to the nearest prototype; neither quantity is constructed from the zero-shot text-embedding baselines it is compared against. For ESC-50 and FSD50K the zero-shot rows are imported from the cited papers (Table 1 caption), and for BBL the zero-shot row is re-run from open-source implementations; using an external baseline is not circular, though the uncontrolled comparison is a correctness risk. The FSD50K support-file selection with 'least class overlap' (Section 4.2) and the selection of 'the best K' for MI variants (Table 1 caption) reflect evaluation choices that could inflate results, but they are overfitting concerns rather than self-referential reasoning. No load-bearing step is justified by self-citation, no fitted parameter is renamed as a prediction, and no result is defined in terms of its own output. The central claim that audio-derived class representations improve over zero-shot text embeddings is therefore self-contained and not circular.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new entities. The only fitted parameter is the MI feature-count ratio K, selected on the evaluation set. Key assumptions are the comparability of external zero-shot baselines and the representativeness of the selected FSD50K support files, both of which affect the strength of the reported conclusions.

free parameters (1)

MI feature-count ratio K = best K selected per dataset/method on the evaluation set (values from {1/2, 1, 2, 4, 8, 16, 32})
K controls the embedding dimension after mutual-information selection. The authors select the K that maximizes accuracy/mAP on the test set (Table 1 caption, Figure 3), which is a hyperparameter fitted to the evaluation data.

assumptions (4)

domain assumption CLAP audio and text embeddings are fixed and provide a meaningful shared embedding space for audio classification.
The entire few-shot pipeline relies on the pretrained CLAP encoder (Section 4.2) without re-training or verifying that its embedding space is suitable for these datasets.
domain assumption Evaluation scores of external zero-shot baselines (CLAP22, PENGI, CLAP23) were obtained under comparable conditions to the few-shot experiments.
For ESC-50 and FSD50K, zero-shot numbers are taken from the respective papers (Table 1 caption) rather than re-run with the same preprocessing and prompts, so comparability is assumed.
standard math LDA performance is degraded by the curse of dimensionality for 1024-dimensional embeddings.
The paper uses this to explain why LDA underperforms averaging (Section 5.1), but it is an explanatory assumption, not a derived result.
domain assumption The FSD50K few-shot support sets with least class overlap are representative of practical few-shot conditions.
Section 4.2 selects least-overlap files, which may not represent random or difficult samples, biasing results.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Improving Audio Classification by Transitioning from Zero- to Few-Shot." pith.science (2026). https://pith.science/paper/IK3XMEP5

@misc{pith2026250720036,
  author       = {Pith},
  title        = {Pith review of: Improving Audio Classification by Transitioning from Zero- to Few-Shot},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IK3XMEP5}},
  note         = {Machine review of arXiv:2507.20036}
}

read the original abstract

State-of-the-art audio classification often employs a zero-shot approach, which involves comparing audio embeddings with embeddings from text describing the respective audio class. These embeddings are usually generated by neural networks trained through contrastive learning to align audio and text representations. Identifying the optimal text description for an audio class is challenging, particularly when the class comprises a wide variety of sounds. This paper examines few-shot methods designed to improve classification accuracy beyond the zero-shot approach. Specifically, audio embeddings are grouped by class and processed to replace the inherently noisy text embeddings. Our results demonstrate that few-shot classification typically outperforms the zero-shot baseline.

Figures

Figures reproduced from arXiv: 2507.20036 by the authors.

**Figure 1.** Mean and standard deviation of FS|Ec | AVG and FS|Ec | LDA using between 2 and 15 training samples, using 30 runs for each |Ec | value. CLAP23 (FT) is the fine-tuned CLAP model for that specific data set. mAP (FSD50K). A summary of the results is given in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** Confusion matrix for BBL. On the left, we see the comparison between the true label and using base CLAP23. On the right, we see a much stronger diagonal trend and an improvement in most classes when using FS|Ec | AVG 0.5 1 2 4 8 16 32 0.4 0.6 0.8 1 K Accuracy/mAP FS|Ec | MI+AVG BBL (|Ec |=10) BBL (|Ec |=20) BBL (|Ec |=50) FSD50K (|Ec |=10) FSD50K (|Ec |=20) FSD50K (|Ec |=50) ESC-50 (|Ec |=20) 1 2 4 8 16 32 K FS|Ec |… view at source ↗

**Figure 3.** Accuracy as we increase K. Solid horizontal lines indicate the baseline, CLAP23. The left plot shows continual improvements as we use MI as a feature selector before averaging, while we see a more complex relationship when using MI as an intermediary step before LDA. To investigate the number of required features, we plot the accuracy/mAP over K in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages

[1]

Competitions like DCASE [1] and accompanying datasets like TUT [2] and CochlScene [3] showcase the importance and interest in acous- tic scene classification and event detection

Introduction Identification and classification of sound events play a vital role in understanding acoustic environments. Competitions like DCASE [1] and accompanying datasets like TUT [2] and CochlScene [3] showcase the importance and interest in acous- tic scene classification and event detection. Practitioners work- ing on audio machine learning rely on...

work page
[2]

Improving Audio Classification by Transitioning from Zero- to Few-Shot

Fundamentals We consider a dataset X = {xn | n = 1, . . . , N}, consisting of N discrete audio signals xn ∈ RTn, where Tn ∈ N speci- fies the number of samples in each signal. Correspondingly, we define a set of labels Y = {yn | n = 1, . . . , N}, with yn ∈ C arXiv:2507.20036v1 [cs.SD] 26 Jul 2025 representing the class label of the n-th audio recording, ...

work page Pith review arXiv 2025
[3]

Few-Shot Audio Classification Methods In many practical scenarios, a small number of audio samples for all classes already exists or can be gathered with minimal effort. Extracting audio embeddings from such a development set and constructing ec from them in a few-shot manner instead of text leads to an improved classification, as we can see in Sec- tion ...

work page
[4]

Data The performance evaluation is based on three datasets: BBL (internal), ESC-50, and FSD50K

Experimental Setup 4.1. Data The performance evaluation is based on three datasets: BBL (internal), ESC-50, and FSD50K. These datasets provide a di- verse range of sound events and evaluation scenarios. BBL consists of 3600 audio signals divided equally among 24 hu- manly annotated sound classes. The audio originates from Au- dioSet [30]. The classes cove...

work page 2000
[5]

For ESC-50 and FSD50K, results are obtained from the respective papers

Performance Evaluation We compared different zero-shot and FS methods to each other using their classification accuracy (BBL and ESC-50) or the Table 1: Top shows zero-shot baselines, bottom shows FS methods. For ESC-50 and FSD50K, results are obtained from the respective papers. For BBL, results are obtained using open-source implementations. For methods...

work page
[6]

Our few-shot approach, uti- lizing a small number of audio samples, consistently outper- formed zero-shot classifiers on different datasets by 2 to 10 % points in accuracy

Conclusion In this study, we improved audio classification by transition- ing from zero-shot to few-shot methods, addressing the limi- tations of noisy text embeddings. Our few-shot approach, uti- lizing a small number of audio samples, consistently outper- formed zero-shot classifiers on different datasets by 2 to 10 % points in accuracy. This success is...

work page
[7]

Sound Event Detection in the DCASE 2017 Challenge,

A. Mesaros, A. Diment, B. Elizalde, T. Heittola, E. Vincent, B. Raj, and T. Virtanen, “Sound Event Detection in the DCASE 2017 Challenge,”IEEE/ACM Trans. Audio Speech Lang. Process., vol. 27, no. 6, pp. 992–1006, 2019

work page 2017
[8]

TUT database for acoustic scene classification and sound event detection,

A. Mesaros, T. Heittola, and T. Virtanen, “TUT database for acoustic scene classification and sound event detection,” in Eur. Signal Process. Conf., 2016, pp. 1128–1132

work page 2016

Show all 38 references

[9]

CochlScene: Acquisition of acoustic scene data using crowdsourcing,

I. Y . Jeong and J. Park, “CochlScene: Acquisition of acoustic scene data using crowdsourcing,” inProc. Asia-Pacific Signal Inf. Process. Assoc. Annu. Summit Conf. APSIPA ASC, 2022, pp. 17– 21

2022
[10]

A dataset and taxonomy for urban sound research,

J. Salamon, C. Jacoby, and J. P. Bello, “A dataset and taxonomy for urban sound research,” in MM - Proc. ACM Conf. Multimed., ser. MM ’14, 2014, pp. 1041–1044

2014
[11]

AI-Based Acoustic Surveillance System for Smart Cities,

J. Holub and J. Svatos, “AI-Based Acoustic Surveillance System for Smart Cities,” in 8th IEEE Int. Forum Res. Technol. Soc. Ind. Innov. RTSI 2024 - Proceeding, 2024, pp. 295–299

2024
[12]

Acoustic monitoring in- tegrated with electronic care planning,

NHS Transformation Directorate, “Acoustic monitoring in- tegrated with electronic care planning,” https://transform. england.nhs.uk/ai-lab/explore-all-resources/understand-ai/ acoustic-monitoring-integrated-electronic-care-planning/, 2020

2020
[13]

Passive acoustic monitoring in ecology and conserva- tion,

E. Browning, R. Gibb, P. Glover-Kapfer, K. E. Jones, G. Billing- ton, Z. Burivalova, D. Clink, J. De Ridder, J. Halls, T. Hast- ings, D. Jacoby, A. Kalan, A. Kershenbaum, S. Linke, S. Lucas, R. Machado, P. Owens, C. Sutter, P. Trethowan, R. Whytock, and P. Wrege, “Passive acou...

2017
[14]

SECP: A Speech Enhancement-Based Curation Pipeline for Scalable Acquisition of Clean Speech,

A. Sabra, C. Wronka, M. Mao, and S. Hijazi, “SECP: A Speech Enhancement-Based Curation Pipeline for Scalable Acquisition of Clean Speech,” inICASSP , IEEE Int. Conf. Acoust. Speech Sig- nal Process. - Proc., 2024, pp. 11 981–11 985

2024
[15]

Environmental sound classification with convolu- tional neural networks,

K. J. Piczak, “Environmental sound classification with convolu- tional neural networks,” in IEEE Int. Work. Mach. Learn. Signal Process. MLSP, 2015, pp. 1–6

2015
[16]

Ro- bust sound event classification using deep neural networks,

I. McLoughlin, H. Zhang, Z. Xie, Y . Song, and W. Xiao, “Ro- bust sound event classification using deep neural networks,”IEEE Trans. Audio, Speech Lang. Process., vol. 23, no. 3, pp. 540–552, 2015

2015
[17]

DCasenet: An integrated pretrained deep neural network for detecting and clas- sifying acoustic scenes and events,

J. W. Jung, H. J. Shim, J. H. Kim, and H. J. Yu, “DCasenet: An integrated pretrained deep neural network for detecting and clas- sifying acoustic scenes and events,” in ICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2021, pp. 621–625

2021
[18]

Few-shot continual learning for audio classification,

Y . Wang, N. J. Bryan, M. Cartwright, J. P. Bello, and J. Salamon, “Few-shot continual learning for audio classification,” inICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2021, pp. 321–325

2021
[19]

Towards Robust Few-shot Class Incremental Learning in Audio Classification using Con- trastive Representation,

R. Singh, P. Nema, and V . K. Kurmi, “Towards Robust Few-shot Class Incremental Learning in Audio Classification using Con- trastive Representation,” in Interspeech, 2024, pp. 5023–5027

2024
[20]

Multi-label Few-shot Learning for Sound Event Recognition,

K. H. Cheng, S. Y . Chou, and Y . H. Yang, “Multi-label Few-shot Learning for Sound Event Recognition,” in IEEE 21st Int. Work. Multimed. Signal Process. MMSP, 2019, pp. 1–5

2019
[21]

Learning From Taxonomy: Multi-Label Few-Shot Classification for Everyday Sound Recog- nition,

J. Liang, H. Phan, and E. Benetos, “Learning From Taxonomy: Multi-Label Few-Shot Classification for Everyday Sound Recog- nition,” in ICASSP , IEEE Int. Conf. Acoust. Speech Signal Pro- cess. - Proc., 2024, pp. 771–775

2024
[22]

Learning Transferable Visual Models From Natu- ral Language Supervision,

A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agar- wal, G. Sastry, A. Askell, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever, “Learning Transferable Visual Models From Natu- ral Language Supervision,”Proc. Mach. Learn. Res., vol. 139, pp. 8748–8763, 2021

2021
[23]

Wav2Clip: Learning Robust Audio Representations From Clip,

H. H. Wu, P. Seetharaman, K. Kumar, and J. P. Bello, “Wav2Clip: Learning Robust Audio Representations From Clip,” in ICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2022, pp. 4563–4567

2022
[24]

Audioclip: Extend- ing Clip To Image, Text and Audio,

A. Guzhov, F. Raue, J. Hees, and A. Dengel, “Audioclip: Extend- ing Clip To Image, Text and Audio,” in ICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2022, pp. 976–980

2022
[25]

CLAP Learning Audio Concepts from Natural Language Supervision,

B. Elizalde, S. Deshmukh, M. A. Ismail, and H. Wang, “CLAP Learning Audio Concepts from Natural Language Supervision,” in ICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2023

2023
[26]

M2D-CLAP: Masked Modeling Duo Meets CLAP for Learning General-purpose Audio-Language Representation,

D. Niizumi, D. Takeuchi, Y . Ohishi, N. Harada, M. Yasuda, S. Tsubaki, and K. Imoto, “M2D-CLAP: Masked Modeling Duo Meets CLAP for Learning General-purpose Audio-Language Representation,” in Interspeech, 2024, pp. 57–61

2024
[27]

Large-Scale Contrastive Language-Audio Pretrain- ing with Feature Fusion and Keyword-to-Caption Augmentation,

Y . Wu, K. Chen, T. Zhang, Y . Hui, T. Berg-Kirkpatrick, and S. Dubnov, “Large-Scale Contrastive Language-Audio Pretrain- ing with Feature Fusion and Keyword-to-Caption Augmentation,” in ICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2023, pp. 1–5

2023
[28]

Natural Language Supervision For General-Purpose Audio Representations,

B. Elizalde, S. Deshmukh, and H. Wang, “Natural Language Supervision For General-Purpose Audio Representations,” in ICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2024, pp. 336–340

2024
[29]

Audio-Text Models Do Not Yet Leverage Natural Language,

H. H. Wu, O. Nieto, J. P. Bello, and J. Salamon, “Audio-Text Models Do Not Yet Leverage Natural Language,” in ICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2023, pp. 1–5

2023
[30]

Pengi: An Audio Language Model for Audio Tasks,

S. Deshmukh, B. Elizalde, R. Singh, and H. Wang, “Pengi: An Audio Language Model for Audio Tasks,” Adv. Neural Inf. Pro- cess. Syst., vol. 36, 2023

2023
[31]

Audio Flamingo: A Novel Audio Language Model with Few- Shot Learning and Dialogue Abilities,

Z. Kong, A. Goel, R. Badlani, W. Ping, R. Valle, and B. Catanzaro, “Audio Flamingo: A Novel Audio Language Model with Few- Shot Learning and Dialogue Abilities,” in Proc. Mach. Learn. Res., vol. 235, 2024, pp. 25 125–25 148

2024
[32]

N. P. Fa ´ısca, K. I. Kouramas, B. Rustem, and E. N. Pistikopou- los, Dynamic Programming, ser. Rand Corporation research study, 2014, vol. 1-7

2014
[33]

Conceptual and empirical comparison of dimensionality reduction algorithms (PCA, KPCA, LDA, MDS, SVD, LLE, ISOMAP, LE, ICA, t-SNE),

F. Anowar, S. Sadaoui, and B. Selim, “Conceptual and empirical comparison of dimensionality reduction algorithms (PCA, KPCA, LDA, MDS, SVD, LLE, ISOMAP, LE, ICA, t-SNE),” Comput. Sci. Rev., vol. 40, no. C, 2021

2021
[34]

Using Mutual Information for Selecting Features in Supervised Neural Net Learning,

R. Battiti, “Using Mutual Information for Selecting Features in Supervised Neural Net Learning,” IEEE Trans. Neural Networks, vol. 5, no. 4, pp. 537–550, 1994

1994
[35]

A review of feature selection methods based on mutual information,

J. R. Vergara and P. A. Est ´evez, “A review of feature selection methods based on mutual information,” Neural computing and applications, vol. 24, pp. 175–186, 2014

2014
[36]

Audio Set: An ontology and human-labeled dataset for audio events,

J. F. Gemmeke, D. P. Ellis, D. Freedman, A. Jansen, W. Lawrence, R. C. Moore, M. Plakal, and M. Ritter, “Audio Set: An ontology and human-labeled dataset for audio events,” inICASSP , IEEE Int. Conf. Acoust. Speech Signal Process. - Proc., 2017, pp. 776–780

2017
[37]

ESC: Dataset for environmental sound classifica- tion,

K. J. Piczak, “ESC: Dataset for environmental sound classifica- tion,” in MM 2015 - Proc. 2015 ACM Multimed. Conf. , ser. MM ’15, 2015, pp. 1015–1018

2015
[38]

FSD50K: An Open Dataset of Human-Labeled Sound Events,

E. Fonseca, X. Favory, J. Pons, F. Font, and X. Serra, “FSD50K: An Open Dataset of Human-Labeled Sound Events,” IEEE/ACM Trans. Audio Speech Lang. Process., vol. 30, pp. 829–852, 2022

2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Competitions like DCASE [1] and accompanying datasets like TUT [2] and CochlScene [3] showcase the importance and interest in acous- tic scene classification and event detection

Introduction Identification and classification of sound events play a vital role in understanding acoustic environments. Competitions like DCASE [1] and accompanying datasets like TUT [2] and CochlScene [3] showcase the importance and interest in acous- tic scene classification and event detection. Practitioners work- ing on audio machine learning rely on...

work page

[2] [2]

Improving Audio Classification by Transitioning from Zero- to Few-Shot

Fundamentals We consider a dataset X = {xn | n = 1, . . . , N}, consisting of N discrete audio signals xn ∈ RTn, where Tn ∈ N speci- fies the number of samples in each signal. Correspondingly, we define a set of labels Y = {yn | n = 1, . . . , N}, with yn ∈ C arXiv:2507.20036v1 [cs.SD] 26 Jul 2025 representing the class label of the n-th audio recording, ...

work page Pith review arXiv 2025

[3] [3]

Few-Shot Audio Classification Methods In many practical scenarios, a small number of audio samples for all classes already exists or can be gathered with minimal effort. Extracting audio embeddings from such a development set and constructing ec from them in a few-shot manner instead of text leads to an improved classification, as we can see in Sec- tion ...

work page

[4] [4]

Data The performance evaluation is based on three datasets: BBL (internal), ESC-50, and FSD50K

Experimental Setup 4.1. Data The performance evaluation is based on three datasets: BBL (internal), ESC-50, and FSD50K. These datasets provide a di- verse range of sound events and evaluation scenarios. BBL consists of 3600 audio signals divided equally among 24 hu- manly annotated sound classes. The audio originates from Au- dioSet [30]. The classes cove...

work page 2000

[5] [5]

For ESC-50 and FSD50K, results are obtained from the respective papers

Performance Evaluation We compared different zero-shot and FS methods to each other using their classification accuracy (BBL and ESC-50) or the Table 1: Top shows zero-shot baselines, bottom shows FS methods. For ESC-50 and FSD50K, results are obtained from the respective papers. For BBL, results are obtained using open-source implementations. For methods...

work page

[6] [6]

Our few-shot approach, uti- lizing a small number of audio samples, consistently outper- formed zero-shot classifiers on different datasets by 2 to 10 % points in accuracy

Conclusion In this study, we improved audio classification by transition- ing from zero-shot to few-shot methods, addressing the limi- tations of noisy text embeddings. Our few-shot approach, uti- lizing a small number of audio samples, consistently outper- formed zero-shot classifiers on different datasets by 2 to 10 % points in accuracy. This success is...

work page

[7] [7]

Sound Event Detection in the DCASE 2017 Challenge,

A. Mesaros, A. Diment, B. Elizalde, T. Heittola, E. Vincent, B. Raj, and T. Virtanen, “Sound Event Detection in the DCASE 2017 Challenge,”IEEE/ACM Trans. Audio Speech Lang. Process., vol. 27, no. 6, pp. 992–1006, 2019

work page 2017

[8] [8]

TUT database for acoustic scene classification and sound event detection,

A. Mesaros, T. Heittola, and T. Virtanen, “TUT database for acoustic scene classification and sound event detection,” in Eur. Signal Process. Conf., 2016, pp. 1128–1132

work page 2016