arxiv: 2601.09448 · v2 · submitted 2026-01-14 · 💻 cs.SD · cs.AI

One Prompt, Many Sounds: Modeling Listener Variability in LLM-Based Equalization

Ioannis Stylianou , Jon Francombe , Pablo Martinez-Nuevo , Sven Ewan Shepstone , Zheng-Hua Tan This is my paper

Pith reviewed 2026-05-16 14:23 UTC · model grok-4.3

classification 💻 cs.SD cs.AI

keywords LLMaudio equalizationnatural languagelistener variabilityin-context learningparameter-efficient fine-tuningdistributional alignment

0 comments p. Extension

The pith

Large language models can map natural language prompts to audio equalization settings that match population preferences.

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

The paper establishes that LLMs can convert everyday text descriptions of listening contexts into equalization parameters. It collects data from a controlled listening experiment and applies in-context learning plus parameter-efficient fine-tuning to produce settings that align with what groups of listeners prefer. Evaluation uses distributional metrics to show statistically significant gains over random sampling and fixed preset baselines. The approach treats equalization as a conversational task rather than a manual adjustment process. This opens the possibility of audio systems that adapt automatically to mood, location, or social setting without expert intervention.

Core claim

The central claim is that LLMs function as artificial equalizers by mapping natural language text prompts to equalization settings that reliably align with population-preferred values. This is achieved by training on data from a controlled listening experiment using in-context learning and parameter-efficient fine-tuning techniques, which produce statistically significant improvements in distributional alignment compared with random sampling and static preset baselines.

What carries the argument

An LLM that uses in-context learning and parameter-efficient fine-tuning on listening-experiment data to translate natural language prompts into equalization parameters.

If this is right

Audio systems could accept spoken or written descriptions to adjust sound automatically.
Consumer devices could replace manual EQ sliders with conversational control.
Static factory presets could be replaced by prompt-driven adaptation that reflects group taste distributions.
Evaluation methods based on distributional metrics could be applied to other audio control tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same prompting approach might extend to related audio controls such as dynamic range compression or spatial effects.
Individual users could supply a few personal examples to shift the model from population alignment toward personal preference.
Real-time integration with voice assistants would allow EQ changes mid-listening session based on simple descriptions.

Load-bearing premise

The data from the controlled listening experiment sufficiently captures listener variability and generalizes to new natural language prompts and contexts.

What would settle it

Collect listener preferences for a set of prompts never shown during training or fine-tuning, then check whether the model's predicted equalization settings show the same statistically significant distributional alignment improvement over baselines that was reported on the original data.

Figures

Figures reproduced from arXiv: 2601.09448 by Ioannis Stylianou, Jon Francombe, Pablo Martinez-Nuevo, Sven Ewan Shepstone, Zheng-Hua Tan.

**Figure 3.** Figure 3: The filters that comprise the equalization controller. A horizontal [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

**Figure 4.** Figure 4: Outline of RAG and RAG-QA recommendation approaches. The sentence similarity is computed as the dot product between the embedding of the [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

**Figure 5.** Figure 5: Parameter Efficient Fine-Tuning with regression head. The figure on the left illustrates the details of Prefix Tuning and LoRA methods, while the [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗

**Figure 6.** Figure 6: Visualization of listener agreement in the Beosonic space. (a) For [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

**Figure 7.** Figure 7: Example of standard KDE vs reflective KDE. The blue/white points are drawn from a uniform distribution in the [-6,6]x[-6,6] square. With the [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗

**Figure 8.** Figure 8: Example of standard KDE vs reflective KDE. The blue/white points are drawn from a bimodal standard Gaussian distribution with modes [-3,3] and [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗

**Figure 9.** Figure 9: Visual validation of the Kantorovich distance as a performance metric. The plots contrast the Ground Truth user responses (Orange) against two synthetic model predictions: a “Population-Aligned” distribution (Blue) and a “Misaligned” distribution (Red). (a) Simple Prompt: The Red samples are spatially shifted from the consensus. The Kantorovich distance accurately reflects this error with a high value comp… view at source ↗

**Figure 12.** Figure 12: Comparison between proposed ICL and PEFT methods using [PITH_FULL_IMAGE:figures/full_fig_p010_12.png] view at source ↗

**Figure 13.** Figure 13: Comparison between proposed ICL and PEFT methods using [PITH_FULL_IMAGE:figures/full_fig_p010_13.png] view at source ↗

read the original abstract

Conventional audio equalization is a static process that requires manual and cumbersome adjustments to adapt to changing listening contexts (e.g., mood, location, or social setting). In this paper, we introduce a Large Language Model (LLM)-based alternative that maps natural language text prompts to equalization settings. This enables a conversational approach to sound system control. By utilizing data collected from a controlled listening experiment, our models exploit in-context learning and parameter-efficient fine-tuning techniques to reliably align with population-preferred equalization settings. Our evaluation methods, which leverage distributional metrics that capture users' varied preferences, show statistically significant improvements in distributional alignment over random sampling and static preset baselines. These results indicate that LLMs could function as "artificial equalizers," contributing to the development of more accessible, context-aware, and expert-level audio tuning methods.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

LLMs map text prompts to EQ settings with distributional gains over baselines on their listener data, but the experiment details and generalization checks are missing.

read the letter

The main takeaway is that they ran a listening experiment to gather how different people set EQ for the same text prompts, then used that data with in-context learning and parameter-efficient fine-tuning to make LLMs output settings that better match the spread of listener preferences instead of a single target. They report statistically significant improvements on distributional metrics compared to random sampling and fixed presets. The conversational control angle is straightforward and the choice to model variability through distributions rather than averages makes sense for this domain. The work is grounded in actual collected responses rather than purely synthetic or self-referential setups, which gives it a practical edge for consumer audio interfaces. The soft spot is the missing methodological detail. No participant count, prompt selection process, exact statistical tests, or held-out prompt evaluation appears in the abstract, so it is unclear whether the reported gains reflect robust generalization or just in-sample fit. The stress-test note on generalization from the controlled data to unseen prompts and contexts looks like the right place to press. This is for audio engineers or HCI researchers exploring LLM-driven personalization in sound systems. A reader building adaptive audio tools could borrow the prompt-to-distribution framing even if the current results need more validation. It deserves peer review because the idea is coherent, uses real data, and the gaps are fixable with added experiments and stats rather than fundamental flaws.

Referee Report

2 major / 1 minor

Summary. The paper proposes an LLM-based system that maps natural-language prompts describing listening contexts (mood, location, etc.) to audio equalization settings. It collects listener preference data via a controlled experiment, then applies in-context learning and parameter-efficient fine-tuning to produce models that align with population-level preferred EQ distributions; distributional metrics are used to demonstrate statistically significant gains over random sampling and static-preset baselines.

Significance. If the reported alignment improvements prove robust, the work would offer a practical route to conversational, context-aware audio control that captures inter-listener variability without manual parameter adjustment. The distributional evaluation approach is a constructive way to handle preference diversity, but the absence of core experimental metadata prevents any assessment of whether the gains reflect genuine generalization rather than in-sample fitting.

major comments (2)

[Abstract] Abstract: the assertion of 'statistically significant improvements in distributional alignment' supplies no participant count, exact statistical test, data-collection protocol, or metric definitions, rendering the central empirical claim unverifiable from the provided text.
[Evaluation] Evaluation section: no held-out prompt set, prompt-diversity statistics, or cross-context generalization results are reported, leaving the claim that in-context learning and PEFT 'reliably align' with population preferences on arbitrary unseen prompts unsupported.

minor comments (1)

[Methods] Notation for the distributional metrics (e.g., how the preferred EQ distribution is constructed from listener responses) should be defined explicitly before the results are presented.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback, which highlights important issues of verifiability and generalization. We address each major comment below and will revise the manuscript to strengthen these aspects.

read point-by-point responses

Referee: [Abstract] Abstract: the assertion of 'statistically significant improvements in distributional alignment' supplies no participant count, exact statistical test, data-collection protocol, or metric definitions, rendering the central empirical claim unverifiable from the provided text.

Authors: We agree that the abstract is currently too concise and omits key experimental metadata. In the revised manuscript we will expand the abstract to report the participant count from the listening experiment, the exact statistical test and significance level used, a brief description of the data-collection protocol, and explicit definitions of the distributional metrics. This change will make the central claim directly verifiable from the abstract. revision: yes
Referee: [Evaluation] Evaluation section: no held-out prompt set, prompt-diversity statistics, or cross-context generalization results are reported, leaving the claim that in-context learning and PEFT 'reliably align' with population preferences on arbitrary unseen prompts unsupported.

Authors: We acknowledge that the evaluation section does not currently include an explicit held-out prompt split, prompt-diversity statistics, or dedicated cross-context generalization results. While the in-context learning setup is designed to support new prompts, we will revise the evaluation section to add these elements: a held-out prompt set, quantitative prompt-diversity measures, and distributional alignment results on the held-out prompts. This will provide direct evidence for generalization beyond the prompts seen during data collection or fine-tuning. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation grounded in new listener data

full rationale

The paper collects fresh data from a controlled listening experiment and applies in-context learning plus PEFT to map prompts to equalization settings that match the observed population preferences. No step reduces by definition to its own inputs, renames a fitted parameter as a prediction, or relies on a load-bearing self-citation chain; the distributional alignment metrics are computed against the collected responses rather than being tautological. The central claim therefore remains independent of the patterns that would trigger a circularity flag.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Central claim depends on the representativeness of the listening experiment data and the transferability of LLM in-context learning to this audio task; no free parameters or invented entities are described.

axioms (1)

domain assumption LLMs can reliably perform in-context learning to map text prompts to continuous parameter settings
Invoked to justify using prompts and fine-tuning without additional derivation.

pith-pipeline@v0.9.0 · 5450 in / 1111 out tokens · 78949 ms · 2026-05-16T14:23:01.899041+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Our models exploit in-context learning and parameter-efficient fine-tuning techniques to reliably align with population-preferred equalization settings... distributional metrics... Kantorovich distance

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

56 extracted references · 56 canonical work pages · 8 internal anchors

[1]

The netflix recommender system: Algorithms, business value, and innovation,

C. A. Gomez-Uribe and N. Hunt, “The netflix recommender system: Algorithms, business value, and innovation,”ACM Transactions on Management Information Systems (TMIS), vol. 6, no. 4, pp. 1–19, 2015

work page 2015
[2]

Music personalization at spotify,

K. Jacobson, V . Murali, E. Newett, B. Whitman, and R. Yon, “Music personalization at spotify,” inProceedings of the 10th ACM Conference on Recommender Systems, ser. RecSys ’16. New York, NY , USA: Association for Computing Machinery, 2016, p. 373. [Online]. Available: https://doi.org/10.1145/2959100.2959120

work page doi:10.1145/2959100.2959120 2016
[3]

Comparison of digital loudspeaker-equalization tech- niques,

P. Dziechci ´nski, “Comparison of digital loudspeaker-equalization tech- niques,”Archives of Acoustics, vol. 30, no. 2, 2005

work page 2005
[4]

All about audio equalization: Solutions and frontiers,

V . V ¨alim¨aki and J. D. Reiss, “All about audio equalization: Solutions and frontiers,”Applied Sciences, vol. 6, no. 5, p. 129, 2016

work page 2016
[5]

Deep optimization of parametric iir filters for audio equalization,

G. Pepe, L. Gabrielli, S. Squartini, C. Tripodi, and N. Strozzi, “Deep optimization of parametric iir filters for audio equalization,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 30, pp. 1136–1149, 2022

work page 2022
[6]

Designing audio equalization filters by deep neural networks,

G. Pepe, L. Gabrielli, S. Squartini, and L. Cattani, “Designing audio equalization filters by deep neural networks,”Applied Sciences, vol. 10, no. 7, p. 2483, 2020

work page 2020
[7]

End-to-end equalization with convolutional neural networks,

M. Martinez Ramirez, J. Reisset al., “End-to-end equalization with convolutional neural networks,” 2018

work page 2018
[8]

Building a personalized audio equal- izer interface with transfer learning and active learning,

B. Pardo, D. Little, and D. Gergle, “Building a personalized audio equal- izer interface with transfer learning and active learning,” inProceedings of the second international ACM workshop on Music information retrieval with user-centered and multimodal strategies, 2012, pp. 13– 18

work page 2012
[9]

Language and culture,

C. Kramsch, “Language and culture,”AILA review, vol. 27, no. 1, pp. 30–55, 2014

work page 2014
[10]

Statistical analysis of an automated in- situ frequency response optimisation algorithm for active loudspeakers,

A. Goldberg and A. M ¨akivirta, “Statistical analysis of an automated in- situ frequency response optimisation algorithm for active loudspeakers,” inproceedings of the 23rd Conf. Audio Eng. Soc.,(2003 May). Citeseer, 2003

work page 2003
[11]

On the variation and invertibility of room impulse response functions,

J. Mourjopoulos, “On the variation and invertibility of room impulse response functions,”Journal of sound and vibration, vol. 102, no. 2, pp. 217–228, 1985

work page 1985
[12]

Audealize: Crowdsourced audio produc- tion tools,

P. Seetharaman and B. Pardo, “Audealize: Crowdsourced audio produc- tion tools,”Journal of the Audio Engineering Society, vol. 64, no. 9, pp. 683–695, 2016

work page 2016
[13]

Audi- olm: a language modeling approach to audio generation,

Z. Borsos, R. Marinier, D. Vincent, E. Kharitonov, O. Pietquin, M. Shar- ifi, D. Roblek, O. Teboul, D. Grangier, M. Tagliasacchiet al., “Audi- olm: a language modeling approach to audio generation,”IEEE/ACM transactions on audio, speech, and language processing, vol. 31, pp. 2523–2533, 2023

work page 2023
[14]

MusicLM: Generating Music From Text

A. Agostinelli, T. I. Denk, Z. Borsos, J. Engel, M. Verzetti, A. Caillon, Q. Huang, A. Jansen, A. Roberts, M. Tagliasacchiet al., “Musiclm: Generating music from text,”arXiv preprint arXiv:2301.11325, 2023

work page internal anchor Pith review Pith/arXiv arXiv 2023
[15]

Audit: Audio editing by following instructions with latent diffusion models,

Y . Wang, Z. Ju, X. Tan, L. He, Z. Wu, J. Bian, and s. zhao, “Audit: Audio editing by following instructions with latent diffusion models,” inAdvances in Neural Information Processing Systems, A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, Eds., vol. 36. Curran Associates, Inc., 2023, pp. 71 340–71 357. [Online]. Available: https://p...

work page 2023
[16]

Prompt-guided precise audio editing with diffusion models,

M. Xu, C. Li, D. Su, W. Liang, D. Yuet al., “Prompt-guided precise audio editing with diffusion models,”arXiv preprint arXiv:2406.04350, 2024

work page arXiv 2024
[17]

Can large language models predict audio effects parame- ters from natural language?

S. Doh, J. Koo, M. A. Mart ´ınez-Ram´ırez, W.-H. Liao, J. Nam, and Y . Mitsufuji, “Can large language models predict audio effects parame- ters from natural language?”arXiv preprint arXiv:2505.20770, 2025

work page arXiv 2025
[18]

Text2fx: Harnessing clap embeddings for text-guided audio effects,

A. Chu, P. O’Reilly, J. Barnett, and B. Pardo, “Text2fx: Harnessing clap embeddings for text-guided audio effects,” inICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025, pp. 1–5

work page 2025
[19]

An automatic volume control for preserving intelligibility,

F. Felber, “An automatic volume control for preserving intelligibility,” in34th IEEE Sarnoff Symposium. IEEE, 2011, pp. 1–5

work page 2011
[20]

Context-aware recom- mender systems in the music domain: A systematic literature review,

A. Lozano Murciego, D. M. Jim ´enez-Bravo, A. Valera Roman, J. F. De Paz Santana, and M. N. Moreno-Garc ´ıa, “Context-aware recom- mender systems in the music domain: A systematic literature review,” Electronics, vol. 10, no. 13, p. 1555, 2021

work page 2021
[21]

Audio-based age and gender identification to enhance the recommendation of tv content,

S. E. Shepstone, Z.-H. Tan, and S. H. Jensen, “Audio-based age and gender identification to enhance the recommendation of tv content,” IEEE Transactions on Consumer Electronics, vol. 59, no. 3, pp. 721– 729, 2013

work page 2013
[22]

Beosound / beovision theatre: Technical sound guide,

B. . O. A/S, “Beosound / beovision theatre: Technical sound guide,” 2023

work page 2023
[23]

Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone

M. Abdinet al., “Phi-3 technical report: A highly capable language model locally on your phone,” 2024. [Online]. Available: https://arxiv.org/abs/2404.14219

work page internal anchor Pith review Pith/arXiv arXiv 2024
[24]

Language mod- els are few-shot learners,

T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askellet al., “Language mod- els are few-shot learners,”Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020

work page 1901
[25]

Automatic text clustering for audio attribute elicitation experiment responses,

J. Francombe, T. Brookes, and R. Mason, “Automatic text clustering for audio attribute elicitation experiment responses,” inAES 143rd Convention. Audio Engineering Society, 2017

work page 2017
[26]

Retrieval- augmented generation for knowledge-intensive nlp tasks,

P. Lewis, E. Perez, A. Piktus, F. Petroni, V . Karpukhin, N. Goyal, H. K ¨uttler, M. Lewis, W.-t. Yih, T. Rockt ¨aschelet al., “Retrieval- augmented generation for knowledge-intensive nlp tasks,”Advances in neural information processing systems, vol. 33, pp. 9459–9474, 2020

work page 2020
[27]

Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

N. Reimers and I. Gurevych, “Sentence-bert: Sentence embeddings using siamese bert-networks,” 2019. [Online]. Available: https: //arxiv.org/abs/1908.10084

work page internal anchor Pith review Pith/arXiv arXiv 2019
[28]

Sentence-t5: Scalable sentence encoders from pre-trained text-to-text models,

J. Ni, G. H. Abrego, N. Constant, J. Ma, K. Hall, D. Cer, and Y . Yang, “Sentence-t5: Scalable sentence encoders from pre-trained text-to-text models,” inFindings of the association for computational linguistics: ACL 2022, 2022, pp. 1864–1874

work page 2022
[29]

Clap learning audio concepts from natural language supervision,

B. Elizalde, S. Deshmukh, M. Al Ismail, and H. Wang, “Clap learning audio concepts from natural language supervision,” inICASSP 2023- 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023, pp. 1–5

work page 2023
[30]

Genqa: Generating millions of instructions from a handful of prompts,

J. Chen, R. Qadri, Y . Wen, N. Jain, J. Kirchenbauer, T. Zhou, and T. Goldstein, “Genqa: Generating millions of instructions from a handful of prompts,” 2024. [Online]. Available: https://arxiv.org/abs/2406.10323

work page arXiv 2024
[31]

On the effectiveness of parameter-efficient fine-tuning,

Z. Fu, H. Yang, A. M.-C. So, W. Lam, L. Bing, and N. Collier, “On the effectiveness of parameter-efficient fine-tuning,” 2022. [Online]. Available: https://arxiv.org/abs/2211.15583

work page arXiv 2022
[32]

Parameter-efficient fine-tuning of large- scale pre-trained language models,

N. Ding, Y . Qin, G. Yang, F. Wei, Z. Yang, Y . Su, S. Hu, Y . Chen, C.-M. Chan, W. Chenet al., “Parameter-efficient fine-tuning of large- scale pre-trained language models,”Nature Machine Intelligence, vol. 5, no. 3, pp. 220–235, 2023

work page 2023
[33]

Lora: Low-rank adaptation of large language models

E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, W. Chenet al., “Lora: Low-rank adaptation of large language models.” ICLR, vol. 1, no. 2, p. 3, 2022

work page 2022
[34]

Parameter-efficient transfer SUBMITTED TO IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING 12 learning for nlp,

N. Houlsby, A. Giurgiu, S. Jastrzebski, B. Morrone, Q. de Laroussilhe, A. Gesmundo, M. Attariyan, and S. Gelly, “Parameter-efficient transfer SUBMITTED TO IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING 12 learning for nlp,” 2019. [Online]. Available: https://arxiv.org/abs/1902. 00751

work page 2019
[35]

Adapterhub: A framework for adapting transformers,

J. Pfeiffer, A. R ¨uckl´e, C. Poth, A. Kamath, I. Vuli ´c, S. Ruder, K. Cho, and I. Gurevych, “Adapterhub: A framework for adapting transformers,”

work page
[36]

Available: https://arxiv.org/abs/2007.07779

[Online]. Available: https://arxiv.org/abs/2007.07779

work page arXiv 2007
[37]

Language models are unsupervised multitask learners,

A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, I. Sutskeveret al., “Language models are unsupervised multitask learners,”OpenAI blog, vol. 1, no. 8, p. 9, 2019

work page 2019
[38]

The Power of Scale for Parameter-Efficient Prompt Tuning

B. Lester, R. Al-Rfou, and N. Constant, “The power of scale for parameter-efficient prompt tuning,” 2021. [Online]. Available: https://arxiv.org/abs/2104.08691

work page internal anchor Pith review Pith/arXiv arXiv 2021
[39]

DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition

J. Donahue, Y . Jia, O. Vinyals, J. Hoffman, N. Zhang, E. Tzeng, and T. Darrell, “Decaf: A deep convolutional activation feature for generic visual recognition,” 2013. [Online]. Available: https: //arxiv.org/abs/1310.1531

work page internal anchor Pith review Pith/arXiv arXiv 2013
[40]

Cross-attention is all you need: Adapting pretrained transformers for machine translation,

M. Gheini, X. Ren, and J. May, “Cross-attention is all you need: Adapting pretrained transformers for machine translation,” 2021. [Online]. Available: https://arxiv.org/abs/2104.08771

work page arXiv 2021
[41]

Scaling down to scale up: A guide to parameter- efficient fine-tuning.arXiv preprint arXiv:2303.15647, 2023

V . Lialin, V . Deshpande, X. Yao, and A. Rumshisky, “Scaling down to scale up: A guide to parameter-efficient fine-tuning,” 2024. [Online]. Available: https://arxiv.org/abs/2303.15647

work page arXiv 2024
[42]

Prefix-Tuning: Optimizing Continuous Prompts for Generation

X. L. Li and P. Liang, “Prefix-tuning: Optimizing continuous prompts for generation,” 2021. [Online]. Available: https://arxiv.org/abs/2101.00190

work page internal anchor Pith review Pith/arXiv arXiv 2021
[43]

Interpolating between Optimal Transport and MMD using Sinkhorn Divergences

J. Feydy, T. S ´ejourn´e, F.-X. Vialard, S. ichi Amari, A. Trouv ´e, and G. Peyr ´e, “Interpolating between optimal transport and mmd using sinkhorn divergences,” 2018. [Online]. Available: https://arxiv.org/abs/ 1810.08278

work page internal anchor Pith review Pith/arXiv arXiv 2018
[44]

Mathematical methods of organizing and planning production,

L. V . Kantorovich, “Mathematical methods of organizing and planning production,”Management science, vol. 6, no. 4, pp. 366–422, 1960

work page 1960
[45]

Long history of the monge-kantorovich transportation problem,

A. Vershik, “Long history of the monge-kantorovich transportation problem,”The Mathematical Intelligencer, vol. 35, 12 2013

work page 2013
[46]

Wasserstein GAN

M. Arjovsky, S. Chintala, and L. Bottou, “Wasserstein gan,” 2017. [Online]. Available: https://arxiv.org/abs/1701.07875

work page internal anchor Pith review Pith/arXiv arXiv 2017
[47]

A multidimensional version of the kolmogorov–smirnov test,

G. Fasano and A. Franceschini, “A multidimensional version of the kolmogorov–smirnov test,”Monthly Notices of the Royal Astronomical Society, vol. 225, no. 1, pp. 155–170, 1987

work page 1987
[48]

Direct preference optimization: Your language model is secretly a reward model,

R. Rafailov, A. Sharma, E. Mitchell, C. D. Manning, S. Ermon, and C. Finn, “Direct preference optimization: Your language model is secretly a reward model,”Advances in neural information processing systems, vol. 36, pp. 53 728–53 741, 2023

work page 2023
[49]

Generative modeling using the sliced wasserstein distance,

I. Deshpande, Z. Zhang, and A. G. Schwing, “Generative modeling using the sliced wasserstein distance,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 3483–3491

work page 2018
[50]

Simple boundary correction for kernel density estimation,

M. C. Jones, “Simple boundary correction for kernel density estimation,” Statistics and computing, vol. 3, pp. 135–146, 1993

work page 1993
[51]

B. W. Silverman,Density estimation for statistics and data analysis. Routledge, 2018

work page 2018
[52]

Incorporating support constraints into nonparametric estimators of densities,

E. F. Schuster, “Incorporating support constraints into nonparametric estimators of densities,”Communications in Statistics-Theory and meth- ods, vol. 14, no. 5, pp. 1123–1136, 1985

work page 1985
[53]

Scott’s rule,

D. W. Scott, “Scott’s rule,”Wiley Interdisciplinary Reviews: Computa- tional Statistics, vol. 2, no. 4, pp. 497–502, 2010

work page 2010
[54]

Variations of box plots,

R. McGill, J. W. Tukey, and W. A. Larsen, “Variations of box plots,” The american statistician, vol. 32, no. 1, pp. 12–16, 1978

work page 1978
[55]

Use of ranks in one-criterion variance analysis,

W. H. Kruskal and W. A. Wallis, “Use of ranks in one-criterion variance analysis,”Journal of the American statistical Association, vol. 47, no. 260, pp. 583–621, 1952

work page 1952
[56]

Multiple comparisons using rank sums,

O. J. Dunn, “Multiple comparisons using rank sums,”Technometrics, vol. 6, no. 3, pp. 241–252, 1964

work page 1964