TSPE: Task-Specific Prompt Ensemble for Improved Zero-Shot Audio Classification

Ashish Seth; Dinesh Manocha; Nishit Anand; Ramani Duraiswami

REVIEW 3 major objections 4 minor 30 references

TSPE: Task-Specific Prompt Ensemble for Improved Zero-Shot Audio Classification

T0 review · 3 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Task-specific prompt ensembles improve zero-shot audio classification by 1.23–16.36% absolute over generic prompts on 12 datasets.

desk verdict Useful idea, but the reported numbers don't survive contact with Table I and the manual curation is effectively per-dataset tuning. read the letter →

arxiv 2501.00398 v2 pith:M6VYR4ZH submitted 2024-12-31 cs.SD cs.AIcs.CLcs.LGeess.AS

Nishit Anand , Ashish Seth , Ramani Duraiswami , Dinesh Manocha This is my paper · ORCID

classification cs.SDcs.AIcs.CLcs.LGeess.AS

keywords zero-shotaudioclassificationpromptensemblehardpromptingaudio-languagemodelCLAPtask-specificpromptsaudio-textalignmentGPT-4generation

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 is trying to establish that the wording of prompts, not the model, is often the bottleneck in zero-shot audio classification. Its method, TSPE (Task-Specific Prompt Ensemble), replaces generic prompts like "sound of a " with a short set of task-specific prompts that add sound attributes and sources, such as "A loud sound of a jackhammer coming from a street." The paper reports that averaging the text embeddings of 20 such prompts per task improves zero-shot accuracy by an absolute 1.23–16.36% over vanilla prompts across 12 datasets and two audio-language models that align sound with text. If this is right, practitioners can improve zero-shot audio classification without any retraining or extra parameters. The paper also reports that a few datasets lose accuracy, so the benefit is not uniform.

What carries the argument

The central mechanism is the task-specific prompt ensemble: for each class label, a set of hard prompts is written as "A <attribute> sound of a <label>", "A sound of a <label> coming from a <source>", and "A <attribute> sound of a <label> can be heard from a <source>". Each prompt is encoded by the text encoder, the resulting embeddings are averaged to form one class-level text representation, and classification is the cosine similarity between that averaged representation and the audio embedding. The ensemble size K=20 is selected by an ablation on VocalSound, and the prompt set itself is filtered by hand to remove GPT-4 hallucinations and semantically mismatched attribute-source pairs. The averaging is what makes the method an ensemble rather than a single prompt.

What would settle it

Run TSPE on a fresh audio dataset whose class labels are withheld from the person who selects the 20 prompts, with K fixed at 20 before evaluation; if the average gain over the generic one-line prompts vanishes, the reported improvements depended on knowing the test labels rather than on the prompt-ensemble idea.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that task-specific hard prompts generated from label information improve audio-text alignment in contrastive audio-language models. The method groups class labels into five task categories, asks GPT-4 to generate 40 candidate prompts per category using sound attributes and sound sources, manually keeps 20 that fit the category, and then averages the text embeddings of those 20 prompts before computing cosine similarity with the audio embedding. On the 12 evaluation datasets, TSPE outperforms vanilla zero-shot prompting for both CLAP-type models tested, with absolute gains from 1.23% to 16.36% and average gains near 2%. The paper notes that performance decreases on some datasets, attributing the drops to prompts that are not linguistically rich enough for those tasks.

Load-bearing premise

The load-bearing premise is that the human selection of 20 prompts per task, and the choice to use 20 prompts based on a test on one of the evaluation datasets, do not amount to quietly tuning the method to the 12 benchmarks.

Editorial extensions

If this is right

A new audio classification task can be handled with no model training: one LLM call generates candidate prompts, a human filters them, and the model's existing text encoder averages the embeddings.
The same 20-prompt set for a task category can be reused across datasets in that category, such as the five instrument-recognition datasets tested.
Because the method only changes text inputs, it can compose with other zero-shot techniques and with any audio-language model that accepts natural-language prompts.
The reported gains are not universal: several datasets lose accuracy, so practitioners should validate TSPE per task before deployment.

Reading between the lines

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

Editorial inference: an automated LLM judge could replace the manual filtering of 40 to 20 prompts; whether it reproduces the human-chosen set would directly test how much of the reported gain is curation.
Editorial inference: because K=20 was tuned on VocalSound, one of the 12 evaluation datasets, the optimal ensemble size is probably task-dependent and should be re-tuned on a held-out split for new tasks.
Editorial inference: the prompt-ensemble recipe is a natural fit for other text-conditioned audio tasks, such as audio retrieval and sound-event localization, but that transfer is not demonstrated in the paper.

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.

Desk Editor's Note

read the letter

Colleague,

Quick take: this paper has a real idea, but it is underevidenced in its current form. The authors apply a known technique—prompt ensembling with LLM-generated descriptors—to zero-shot audio classification, and that is a sensible extension. They also get credit for testing across 12 datasets and two CLAP checkpoints, and for including a limitations section that openly admits the manual filtering and possible bias. That is more honest than many papers.

The problem is that the central claim does not hold up on the paper's own table. Recomputing the per-dataset deltas in Table I gives mean gains of about 2.30% for MS-CLAP'23 and 2.10% for MS-CLAP'22, not the reported 2.06% and 1.89%. The abstract's "1.23-16.36%" range also silently drops several negative results (Beijing Opera -2.11, NSynth Instrument -2.02, Cochlscene -1.09, VocalSound -1.99, and a zero on SESA). That is not a rounding issue; it is a misrepresentation of the experiments. The paper's own text acknowledges performance decreases, so the abstract overstates.

The deeper concern is the manual curation. The authors generate 40 prompts per task with GPT-4 and then hand-pick 20 using their knowledge of the labels and datasets. That is close to fitting the prompt set to the test benchmark. The K=20 setting is tuned on VocalSound, one of the twelve test datasets. Without released prompts or code, and without a control like randomly selected prompt sets or a generic-template ensemble, there is no way to tell whether the gains come from task-specificness or from selection on the test set. The authors don't compare against a random subset of the 40 prompts, which would have been a cheap and telling baseline.

If the numbers were corrected and the method were evaluated with a held-out prompt-selection rule, this could be a useful practitioner note. As is, it is not ready for publication. A serious referee would catch the arithmetic and would likely ask for exactly the controls that are missing. My recommendation: reject in current form, but invite a revision with corrected aggregates, negative results reported, and a prompt-curation protocol that does not touch the test sets.

Who is this for? People working on zero-shot audio classification or prompt engineering for CLAP-style models. It deserves referee time because the question is relevant and the experiments are broad, but the present evidence is not sufficient.

Referee Report

3 major / 4 minor

Summary. The paper proposes TSPE (Task-Specific Prompt Ensemble), a training-free hard-prompting method for zero-shot audio classification with contrastive audio-language models such as MS-CLAP'22 and MS-CLAP'23. TSPE uses GPT-4 to generate attribute- and source-augmented prompts per task category, manually filters the prompts, and averages the text embeddings of the surviving prompts (an ensemble) before computing audio-text similarity. The authors report experiments on 12 audio classification datasets spanning five task categories, claiming an absolute improvement of 1.23–16.36% over vanilla zero-shot evaluation and average gains of 2.06% (MS-CLAP'23) and 1.89% (MS-CLAP'22). The paper also discusses hyperparameter selection of the ensemble size K=20 on the VocalSound dataset and notes some limitations, including the need for manual filtering of GPT-4 outputs.

Significance. If the central claim were correct, TSPE would provide a simple, training-free way to improve zero-shot audio classification, and the idea of using LLM-generated attributes and sources to enrich CLAP-style prompts is interesting and potentially useful. The paper is clearly written and evaluates on a broad set of datasets. However, the manuscript currently does not establish the claimed improvement: the reported aggregate numbers are inconsistent with the paper's own table, the improvement is not universal across datasets, and the manual, dataset-aware prompt curation together with tuning K on a test dataset undermines the zero-shot generalization claim. These issues are load-bearing, so the contribution as presented is not verified.

major comments (3)

[§V-A, Table I, Abstract] The aggregate results reported in §V-A are not supported by Table I. Recomputing per-dataset deltas from Table I gives an average improvement of 2.30% for MS-CLAP'23 (not 2.06%) and 2.10% for MS-CLAP'22 (not 1.89%). Moreover, the abstract's claimed range '1.23-16.36%' selects only positive deltas; for MS-CLAP'23 four of twelve comparisons are negative (e.g., Beijing Opera -2.11, NSynth Instrument -2.02, Cochlscene -1.09, VocalSound -1.99) and one is zero (SESA), while for MS-CLAP'22 seven of twelve are negative (e.g., Mridangam Stroke -3.97, GTZAN -3.21). Thus the central quantitative claim of consistent improvement is contradicted by the paper's own data.
[§III-B, §III-D, §III-E, §V-B] The prompt selection procedure is manual and dataset-aware: the authors state that they 'manually map these sound attributes and sources to each task category' (§III-B) and that they manually filter 20 of 40 GPT-4-generated prompts for semantic fit (§III-D and §III-E). In addition, the ensemble size K=20 is selected by an ablation on VocalSound (§V-B), which is itself one of the twelve test datasets reported in Table I. This means the method's hyperparameters and prompt sets are chosen using the evaluation data, so the reported gains cannot be attributed to a general training-free procedure. A held-out validation split, a fully automatic prompt selection rule, or at least an ablation showing robustness to the manual curation choices is needed to support the zero-shot claim.
[§V-A and Table I] The paper claims TSPE 'significantly outperforms traditional zero-shot evaluation' (Conclusion), but no statistical significance testing is reported. All results are averages over five runs without standard deviations or confidence intervals, and many per-dataset deltas are negative while the positive average gains are modest (2.30% and 2.10% on the corrected computations). Given the high variance across datasets, the conclusion of significant improvement is not supported by the evidence presented.

minor comments (4)

[§III-B and §VII] The Introduction and §III-B describe prompt generation as 'automatic,' but §III-D, §III-E, and Limitation 1 make clear that manual filtering is essential; the degree of automation should be described consistently.
[References [2] and [3]] References [2] and [3] are identical, but they are used for the Mridangam Stroke and Mridangam Tonic datasets respectively; the correct distinct sources for these datasets appear to be missing.
[Introduction] The text introduces the abbreviation ALEs ('Audio-Language Encoders') and then uses ALMs throughout; the terminology should be unified.
[Throughout] There are several typographical issues, including 'V ocalSound' and 'N synth' with irregular spacing, and the caption of Fig. 2 includes 'M SC LAP' with inconsistent spacing.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: the prompt-count hyperparameter K is tuned on VocalSound, one of the 12 reported test datasets, making that row an evaluation-set fit rather than an independent zero-shot prediction; the rest of the evaluation remains independently testable.

fitted input called prediction [Section V-B (Hyper-Parameter Tuning); Table I, row "Vocal Sound (VS)"]
"We select K = 20 prompts from an initial set of 40 generated by GPT-4, and perform an ablation study to determine the optimal value for K. Specifically, we evaluate the performance of TSPE on the VocalSound dataset with K values of {5, 10, 15, 20, 25, 30 } to examine how the number of prompts affects the performance of Audio-Language Models (ALMs) on zero-shot audio classification."

K is a hyperparameter fitted by evaluating TSPE on VocalSound, and Table I reports VocalSound as one of the 12 evaluation datasets ("Vocal Sound (VS) 80.93 78.94 47.09 61.23"). The VocalSound TSPE numbers are therefore not an out-of-sample prediction: the same dataset used to choose K=20 is later presented as a test result. This is the fitted-input-called-prediction pattern. The circularity is partial because K was tuned on only one dataset/model in the table, and the other 11 datasets plus the MSCLAP'22 model still provide independent evaluations; notably, the chosen K did not improve MSCLAP'23 on VocalSound (80.93 to 78.94).

full rationale

The paper has no mathematical derivation chain, so circularity can only arise from how the reported zero-shot predictions are produced. The one concrete circular element is the hyperparameter K: Section V-B selects K=20 by evaluating TSPE on VocalSound, and Table I then includes VocalSound among the 12 zero-shot evaluation datasets, so that cell is an evaluation-set fit rather than an independent prediction. The manual attribute/source filtering in Sections III-B, III-D, and III-E is not formal circularity: it is human prompt curation based on label semantics and dataset knowledge, which is a validity and generalization limitation rather than a construction that makes the accuracy equal to its input. The paper candidly acknowledges this in Section VII, noting that prompt filtering requires human oversight and that task-specific prompts may introduce biases. There are no load-bearing self-citations or imported uniqueness theorems; the CLAP models and the 11 other datasets provide a checkable, externally grounded evaluation. Separately, the aggregate improvements quoted in Section V-A (2.06% and 1.89%) do not match a recomputation from Table I (approximately 2.30% and 2.10%), and the abstract's range 1.23-16.36% omits negative per-dataset deltas; these are correctness inconsistencies, not circularity, but they further weaken the quantitative claim. Overall, the central prompt-ensemble idea is independently testable, so the circularity is partial rather than total.

Assumptions & free parameters 2 free parameters · 2 assumptions · 0 invented entities

The paper's only 'parameters' are the curated prompt sets and the ensemble size K; no new theoretical entities are introduced.

free parameters (2)

K (number of ensembled prompts) = 20
Selected by ablation on VocalSound (Section V-B); the optimum may be dataset-specific.
Prompt set (20 curated prompts per task) = 20 prompts per task, curated manually from 40 GPT-4 outputs
The specific prompts are hand-filtered using task and dataset knowledge (Sections III-B, III-D, III-E), acting as a fitted component of the method.

assumptions (2)

domain assumption Averaging text embeddings of multiple prompts improves zero-shot classification for contrastive audio-language models.
Used throughout Section III-C; the paper does not justify this beyond citing vision-language practice.
domain assumption Manually curated sound attributes and sources for a task category transfer across datasets in that category.
Assumed in Sections III-B and IV; e.g., prompts designed using one dataset's labels are applied to other datasets in the same group.

how reviews work

0 comments

Cite this review

Pith. "Pith review of TSPE: Task-Specific Prompt Ensemble for Improved Zero-Shot Audio Classification." pith.science (2026). https://pith.science/paper/M6VYR4ZH

@misc{pith2026250100398,
  author       = {Pith},
  title        = {Pith review of: TSPE: Task-Specific Prompt Ensemble for Improved Zero-Shot Audio Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M6VYR4ZH}},
  note         = {Machine review of arXiv:2501.00398}
}

read the original abstract

Audio-language models (ALMs) excel in zero-shot audio classification, a task where models classify previously unseen audio clips at test time by leveraging descriptive natural language prompts. We introduce TSPE (Task-Specific Prompt Ensemble), a simple, training-free hard prompting method that boosts ALEs' zero-shot performance by customizing prompts for diverse audio classification tasks. Rather than using generic template-based prompts like "Sound of a car" we generate context-rich prompts, such as "Sound of a car coming from a tunnel". Specifically, we leverage label information to identify suitable sound attributes, such as "loud" and "feeble", and appropriate sound sources, such as "tunnel" and "street" and incorporate this information into the prompts used by Audio-Language Models (ALMs) for audio classification. Further, to enhance audio-text alignment, we perform prompt ensemble across TSPE-generated task-specific prompts. When evaluated on 12 diverse audio classification datasets, TSPE improves performance across ALMs by showing an absolute improvement of 1.23-16.36% over vanilla zero-shot evaluation.

Figures

Figures reproduced from arXiv: 2501.00398 by the authors.

**Figure 1.** Illustration of TSPE Workflow: We start with a pool of sound attributes and sources to customize a set of prompt templates using GPT-4. Then, we [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** Effect of different number of prompts on MSCLAP’23 for Audio [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 22 canonical work pages

[1]

”A study of instrument-wise onset detection in Beijing opera percussion ensembles.” 2014 ieee international conference on acoustics, speech and signal processing (icassp)

Tian, Mi, et al. ”A study of instrument-wise onset detection in Beijing opera percussion ensembles.” 2014 ieee international conference on acoustics, speech and signal processing (icassp). IEEE, 2014

work page 2014
[3]

Akshay Anantapadmanabhan, Ashwin Bellur, and Hema A. Murthy

work page
[5]

2013 IEEE International Con- ference on Acoustics, Speech and Signal Processing, Vancouver, BC, Canada

Mridangam stroke dataset (1.0). 2013 IEEE International Con- ference on Acoustics, Speech and Signal Processing, Vancouver, BC, Canada. Data set

work page 2013
[6]

”Musical genre classification of audio signals.” IEEE Transactions on speech and audio processing 10.5 (2002): 293-302

Tzanetakis, George, and Perry Cook. ”Musical genre classification of audio signals.” IEEE Transactions on speech and audio processing 10.5 (2002): 293-302

work page 2002
[7]

”Neural audio synthesis of musical notes with wavenet autoencoders.” International Conference on Machine Learning

Engel, Jesse, et al. ”Neural audio synthesis of musical notes with wavenet autoencoders.” International Conference on Machine Learning. PMLR, 2017

work page 2017
[8]

Karol J. Piczak. ESC: Dataset for Environmental Sound Classification. In Proceedings of the 23rd Annual ACM Conference on Multimedia, pages 1015–1018. ACM Press

work page
[9]

”CochlScene: Acquisition of acoustic scene data using crowdsourcing.” 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC)

Jeong, Il-Young, and Jeongsoo Park. ”CochlScene: Acquisition of acoustic scene data using crowdsourcing.” 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC). IEEE, 2022

work page 2022
[10]

”A dataset and taxonomy for urban sound research.” Proceedings of the 22nd ACM international conference on Multimedia

Salamon, Justin, Christopher Jacoby, and Juan Pablo Bello. ”A dataset and taxonomy for urban sound research.” Proceedings of the 22nd ACM international conference on Multimedia. 2014

work page 2014

Show all 30 references

[11]

Annamaria Mesaros, Toni Heittola, and Tuomas Virtanen. 2017b. Tut acoustic scenes 2017, development dataset. Data set

2017
[12]

”V ocalsound: A dataset for improving human vocal sounds recognition.” ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Gong, Yuan, Jin Yu, and James Glass. ”V ocalsound: A dataset for improving human vocal sounds recognition.” ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022

2022
[13]

Tito Spadini. 2019. Sound events for surveillance applications (1.0.0). Data set

2019
[14]

”A good prompt is worth millions of parameters: Low-resource prompt-based learning for vision-language models.” arXiv preprint arXiv:2110.08484 (2021)

Jin, Woojeong, et al. ”A good prompt is worth millions of parameters: Low-resource prompt-based learning for vision-language models.” arXiv preprint arXiv:2110.08484 (2021)

2021 arXiv
[15]

”Gpt-4 technical report.” arXiv preprint arXiv:2303.08774 (2023)

Achiam, Josh, et al. ”Gpt-4 technical report.” arXiv preprint arXiv:2303.08774 (2023)

2023 arXiv
[16]

”Learning transferable visual models from natural language supervision.” International conference on machine learning

Radford, Alec, et al. ”Learning transferable visual models from natural language supervision.” International conference on machine learning. PMLR, 2021

2021
[17]

”Learning to prompt for vision-language models.” International Journal of Computer Vision 130.9 (2022): 2337-2348

Zhou, Kaiyang, et al. ”Learning to prompt for vision-language models.” International Journal of Computer Vision 130.9 (2022): 2337-2348

2022
[18]

”Audioclip: Extending clip to image, text and au- dio.” ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Guzhov, Andrey, et al. ”Audioclip: Extending clip to image, text and au- dio.” ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022

2022
[19]

”Wav2clip: Learning robust audio representa- tions from clip.” ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Wu, Ho-Hsiang, et al. ”Wav2clip: Learning robust audio representa- tions from clip.” ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022

2022
[20]

”A systematic survey of prompt engineering in large language models: Techniques and applications.” arXiv preprint arXiv:2402.07927 (2024)

Sahoo, Pranab, et al. ”A systematic survey of prompt engineering in large language models: Techniques and applications.” arXiv preprint arXiv:2402.07927 (2024)

2024 arXiv
[21]

”Clap learning audio concepts from natural language supervision.” ICASSP 2023-2023 IEEE International Confer- ence on Acoustics, Speech and Signal Processing (ICASSP)

Elizalde, Benjamin, et al. ”Clap learning audio concepts from natural language supervision.” ICASSP 2023-2023 IEEE International Confer- ence on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023

2023
[22]

”Maple: Multi-modal prompt learn- ing.” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Khattak, Muhammad Uzair, et al. ”Maple: Multi-modal prompt learn- ing.” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023

2023
[23]

”Compa: Addressing the gap in compositional reasoning in audio-language models.” arXiv preprint arXiv:2310.08753 (2023)

Ghosh, Sreyan, et al. ”Compa: Addressing the gap in compositional reasoning in audio-language models.” arXiv preprint arXiv:2310.08753 (2023)

2023 arXiv
[24]

”Hard prompts made easy: Gradient-based discrete optimization for prompt tuning and discovery.” Advances in Neural Information Processing Systems 36 (2024)

Wen, Yuxin, et al. ”Hard prompts made easy: Gradient-based discrete optimization for prompt tuning and discovery.” Advances in Neural Information Processing Systems 36 (2024)

2024
[25]

”A survey of audio classification using deep learning.” IEEE Access (2023)

Zaman, Khalid, et al. ”A survey of audio classification using deep learning.” IEEE Access (2023)

2023
[26]

”Adapting language-audio models as few-shot audio learners.” arXiv preprint arXiv:2305.17719 (2023)

Liang, Jinhua, et al. ”Adapting language-audio models as few-shot audio learners.” arXiv preprint arXiv:2305.17719 (2023)

2023 arXiv
[27]

”Audio-Free Prompt Tuning for Language-Audio Models.” ICASSP 2024-2024 IEEE In- ternational Conference on Acoustics, Speech and Signal Processing (ICASSP)

Li, Yiming, Xiangdong Wang, and Hong Liu. ”Audio-Free Prompt Tuning for Language-Audio Models.” ICASSP 2024-2024 IEEE In- ternational Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024

2024
[28]

”A Survey of Multimodal Large Language Model from A Data-centric Perspective.” arXiv preprint arXiv:2405.16640 (2024)

Bai, Tianyi, et al. ”A Survey of Multimodal Large Language Model from A Data-centric Perspective.” arXiv preprint arXiv:2405.16640 (2024)

2024 arXiv
[29]

”Multimodal large language models: A survey.” 2023 IEEE International Conference on Big Data (BigData)

Wu, Jiayang, et al. ”Multimodal large language models: A survey.” 2023 IEEE International Conference on Big Data (BigData). IEEE, 2023

2023
[30]

”Grounding Multimodal Large Language Models in Actions.” arXiv preprint arXiv:2406.07904 (2024)

Szot, Andrew, et al. ”Grounding Multimodal Large Language Models in Actions.” arXiv preprint arXiv:2406.07904 (2024)

2024 arXiv
[31]

”Efficient multimodal large language models: A survey.” arXiv preprint arXiv:2405.10739 (2024)

Jin, Yizhang, et al. ”Efficient multimodal large language models: A survey.” arXiv preprint arXiv:2405.10739 (2024)

2024
[33]

Carolan, Kilian, Laura Fennelly, and Alan F. Smeaton. ”A Review of Multi-Modal Large Language and Vision Models.” arXiv preprint arXiv:2404.01322 (2024)

2024 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

”A study of instrument-wise onset detection in Beijing opera percussion ensembles.” 2014 ieee international conference on acoustics, speech and signal processing (icassp)

Tian, Mi, et al. ”A study of instrument-wise onset detection in Beijing opera percussion ensembles.” 2014 ieee international conference on acoustics, speech and signal processing (icassp). IEEE, 2014

work page 2014

[2] [3]

Akshay Anantapadmanabhan, Ashwin Bellur, and Hema A. Murthy

work page

[3] [5]

2013 IEEE International Con- ference on Acoustics, Speech and Signal Processing, Vancouver, BC, Canada

Mridangam stroke dataset (1.0). 2013 IEEE International Con- ference on Acoustics, Speech and Signal Processing, Vancouver, BC, Canada. Data set

work page 2013

[4] [6]

”Musical genre classification of audio signals.” IEEE Transactions on speech and audio processing 10.5 (2002): 293-302

Tzanetakis, George, and Perry Cook. ”Musical genre classification of audio signals.” IEEE Transactions on speech and audio processing 10.5 (2002): 293-302

work page 2002

[5] [7]

”Neural audio synthesis of musical notes with wavenet autoencoders.” International Conference on Machine Learning

Engel, Jesse, et al. ”Neural audio synthesis of musical notes with wavenet autoencoders.” International Conference on Machine Learning. PMLR, 2017

work page 2017

[6] [8]

Karol J. Piczak. ESC: Dataset for Environmental Sound Classification. In Proceedings of the 23rd Annual ACM Conference on Multimedia, pages 1015–1018. ACM Press

work page

[7] [9]

”CochlScene: Acquisition of acoustic scene data using crowdsourcing.” 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC)

Jeong, Il-Young, and Jeongsoo Park. ”CochlScene: Acquisition of acoustic scene data using crowdsourcing.” 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC). IEEE, 2022

work page 2022

[8] [10]

”A dataset and taxonomy for urban sound research.” Proceedings of the 22nd ACM international conference on Multimedia

Salamon, Justin, Christopher Jacoby, and Juan Pablo Bello. ”A dataset and taxonomy for urban sound research.” Proceedings of the 22nd ACM international conference on Multimedia. 2014

work page 2014