Continuous Autoregressive Models with Noise Augmentation Avoid Error Accumulation

George Fazekas; Javier Nistal; Marco Pasini; Stefan Lattner

REVIEW 4 major objections 5 minor 23 references

Continuous Autoregressive Models with Noise Augmentation Avoid Error Accumulation

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Training on noise-augmented embeddings prevents error accumulation in continuous autoregressive models.

desk verdict Noise augmentation for continuous AR training is a genuine, clean idea with solid controlled comparisons, but the evidence base is thin enough that the headline claims outrun the measurements. read the letter →

arxiv 2411.18447 v1 pith:E4HUEHIC submitted 2024-11-27 cs.LG cs.AIcs.SDeess.AS

Marco Pasini , Javier Nistal , Stefan Lattner , George Fazekas This is my paper · ORCID

classification cs.LGcs.AIcs.SDeess.AS

keywords continuousautoregressivemodelserroraccumulationnoiseaugmentationrectifiedflowaudiogenerationlatentembeddingscausaltransformerFréchetDistance

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that continuous autoregressive models degrade on long sequences because errors in predicted embeddings compound, and that this can be prevented by training the model on noisy mixtures of clean embeddings and Gaussian noise. The proposed augmentation, $\tilde{x}_t = k_t \varepsilon + (1-k_t) x_t$ with $k_t \sim U(0,1)$, teaches the model to condition on inputs that resemble its own erroneous predictions. A small amount of inference-time noise further stabilizes generation. On musical audio, the method reports FAD 0.405 and FADacc 0.394, improving on all reported autoregressive and non-autoregressive baselines and showing no degradation over extended sequences.

What carries the argument

The load-bearing object is the noisy-mixture input $\tilde{x}_t = k_t \varepsilon + (1-k_t) x_t$, where $k_t \sim U(0,1)$ is sampled per timestep and hidden from the backbone. This forces the causal transformer backbone to act discriminatively, distinguishing real signals from error-like perturbations so that its conditioning $z_t$ stays informative. The sampler is trained with the Rectified Flow objective $L = \mathbb{E}_t[\|v_t - \text{Sampler}(y_t|\sigma_t, z_t)\|^2]$, where $v_t = x_t - \varepsilon$ and $\sigma_t$ follows a lognormal schedule. At inference, adding the fixed noise level $k_{\text{inf}} = 0.02$ to each generated embedding before re-feeding reduces the mismatch between training perturbations and actual prediction errors.

What would settle it

Record the actual prediction errors $e_t = \hat{x}_t - x_t$ from a trained CAM on long generations and compare their distribution and temporal correlation with the assumed model $k_t \varepsilon + (1-k_t)x_t$, $k_t \sim U(0,1)$; if the empirical errors are correlated across timesteps or concentrated at certain mixing levels, training with the uniform schedule should underperform an alternative that uses the measured error distribution.

Watch

Extended reading notes

Core claim

The paper sets out to show that the error-accumulation failure of purely autoregressive models on continuous embeddings can be removed by a simple training-time noise augmentation. The authors model a generated embedding as $\tilde{x}_t = k_t \varepsilon + (1-k_t) x_t$ with $\varepsilon \sim \mathcal{N}(0,I)$ and $k_t$ drawn uniformly from $[0,1]$, and feed these noisy mixtures to a causal backbone whose representation $z_t$ conditions a Rectified Flow sampler. The backbone is not told the value of $k_t$, so it must learn to separate real content from error-like signal. During inference, each generated embedding is lightly re-noised with $k_{\text{inf}} = 0.02$ before being fed back. On musical audio, this yields FAD 0.405 and FADacc 0.394, with the long-sequence score matching or beating the first-sequence score and surpassing all reported baselines.

Load-bearing premise

The method rests on the assumption that inference errors look like convex combinations of clean embeddings and isotropic Gaussian noise with uniform mixing levels, which the paper does not directly validate.

Editorial extensions

If this is right

Causal GPT-style decoding on continuous embeddings becomes viable without masking or bidirectional attention, so key-value caching and other decoding optimizations can be used.
The augmentation transfers to other continuous autoregressive formulations: adding it to GIVT with 32 modes lowers its FAD from 0.865 to 0.514 and its FADacc from 0.931 to 0.511.
Because FADacc does not exceed FAD in CAM, the model does not show the usual quality degradation over the second 10 seconds of generation.
The inference-time noise level $k_{\text{inf}}$ acts as a tunable robustness knob, and at $k_{\text{inf}} = 0.02$ the long-sequence score actually improves relative to the first 10 seconds.
Purely autoregressive continuous generation can be efficient enough for real-time interactive audio applications, since the sampler is a shallow MLP and denoising cost is small relative to the backbone's forward pass.

Reading between the lines

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

Editorial inference: the uniform schedule $k_t \sim U(0,1)$ covers all mixing levels evenly, but generation errors may concentrate at small $k$ early in a sequence and grow with context; a curriculum that ties $k_t$ to positional or measured error statistics might improve training efficiency.
Editorial inference: because the paper does not measure the actual distribution of inference errors, the strongest test is to record $e_t = \hat{x}_t - x_t$ and compare its marginals and autocorrelation to the assumed Gaussian-mixture form; if the match is poor, an error-adaptive augmentation should beat the uniform schedule.
Editorial inference: the finding that FADacc is lower than FAD may reflect the full 10-second context available after the first block rather than a general absence of error accumulation; isolating the context-length effect would clarify how much of the gain comes from noise augmentation alone.
Editorial inference: noise augmentation may act as a regularizer that stops the backbone from over-trusting its own previous outputs, a mechanism that could apply to any conditional generative model whose inputs are model-generated.

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: the core idea—train a causal transformer on noisy convex mixtures of input embeddings to simulate its own inference-time errors—is genuine and clean. The controlled comparisons use the same architecture and parameter counts, and the method beats MAR, MAR-RF, and GIVT variants on FAD, with a consistent gain and even lower FADacc. I also credit the GIVT+noise ablation: showing the same training trick helps a different AR baseline makes the mechanism less like a fluke. The writing is clear.

The soft spots start with the evidence base. No error bars or significance tests; FAD differences of 0.03–0.04 are modest, and with a single internal dataset and one evaluator, I can't tell if the gap is stable. Second, and more load-bearing, the assumed error model in Eq. (4) is never verified. They posit that accumulated errors are convex mixtures of clean embeddings and isotropic Gaussian noise, with mixing level uniform in [0,1]. But real errors are generated autoregressively—they are temporally correlated, state-dependent, and likely non-Gaussian. The stress-test note is right: training with independent per-step injected noise can make the Backbone tolerate arbitrary corruption, but that's not the same as robustness to the specific compounding errors at inference. A distributional check between injected noise and measured prediction errors would close the gap; without it, the central mechanism rests on an equivalence that is plausible but untested.

Third, kinf=0.02 is selected on the same FAD/FADacc metrics that are then reported as evidence. That's a selection effect, not fatal, but it inflates confidence in the stated gains. Fourth, the FADacc < FAD result is confounded by context length—the first 10s have partial context, the second 10s have full context—and the paper admits this, so the 'no degradation' claim is weaker than advertised.

What's genuinely new: applying scheduled-sampling-style noise augmentation to continuous-embedding autoregressive models, with the specific trick of not telling the Backbone the noise level. The omission of scheduled sampling and related robustness literature is a minor citation gap, not a flaw in the work.

Bottom line: this is a solid workshop-level result with a load-bearing assumption that needs checking. The idea will probably be built on. Send it to serious referees, but require the error-distribution validation, error bars, and ideally an external dataset or code release before the claims are accepted as stated.

Referee Report

4 major / 5 minor

Summary. The paper proposes Continuous Autoregressive Models (CAMs) trained with a noise-augmentation scheme to mitigate error accumulation when generating sequences of continuous embeddings in a purely autoregressive (GPT-style) manner. During training, input embeddings are corrupted as x̃_t = k_t ε + (1−k_t) x_t with k_t ∼ U(0,1) (Eq. 4), and at inference a small fixed noise level k_inf is added to generated embeddings before feeding them back to the backbone. Experiments on an internal music-audio latent dataset measure Frechet Audio Distance (FAD) on the first 10 seconds and FADacc on the next 10 seconds. The authors report FAD 0.405 and FADacc 0.394 for CAM, outperforming GIVT, MAR, and a non-autoregressive Rectified Flow baseline, and they show that GIVT also improves when trained with the proposed noise augmentation. The central claims are that noise augmentation makes continuous autoregressive models robust to error accumulation and that CAM preserves quality over extended sequences.

Significance. If the central claims hold, the paper makes a useful and practical contribution: it offers a simple, architecture-agnostic training recipe that enables GPT-style autoregressive generation on continuous embeddings, potentially unlocking KV-cache style efficiency and interactive audio generation. The study has notable strengths: the controlled comparisons use the same backbone architecture and roughly matched parameter counts across baselines, the proposed method is simple and easy to reproduce, and the transfer of the noise-augmentation benefit to GIVT (Fig. 2c) suggests the mechanism is not idiosyncratic to one model. However, the significance is currently limited by three load-bearing gaps: the assumed error model in Eq. (4) is not validated against actual inference errors, the inference noise level k_inf is tuned on the same evaluation metrics used for the headline comparison, and the FADacc metric conflates context length with error accumulation. These issues need to be addressed before the empirical claims can be considered robust.

major comments (4)

[Sec. 4, Eq. (4)] The correctness of the training/inference mismatch argument rests on the assumption that accumulated prediction errors at inference are well modeled by x̃_t = k_t ε + (1−k_t) x_t with ε ∼ N(0, I) and independent k_t ∼ U(0,1). This is never validated. Actual autoregressive errors are produced by the model itself, depend on all previous generated embeddings, and are therefore temporally correlated, state-dependent, and not generally isotropic Gaussian. The paper should provide a diagnostic study comparing the distribution of real inference errors (e.g., per-step residual statistics, autocorrelation, Gaussianity tests) with the injected noise model, or at least an ablation that varies the noise model (e.g., correlated noise, non-Gaussian noise) to show the results are not overly sensitive to this choice. Without such evidence, the claimed equivalence between training-time corruption and inference-time errors is an unverified premise.
[Sec. 5, Fig. 2b (Influence of Inference Noise)] The value k_inf = 0.02 used in all subsequent experiments is selected by the same FAD and FADacc metrics on which the final comparison is based. Because the evaluation set is also used for this model selection, the reported FAD/FADacc of 0.405/0.394 may be optimistically biased. The authors should either tune k_inf on a separate validation set and report the resulting test metrics, or provide error bars and significance tests across multiple independent evaluation draws. Additionally, the paper reports averages over 5 evaluations but no standard deviations; given that the difference between CAM and the best baseline (FAD 0.405 vs 0.442) is small, variance estimates are essential to establish that the improvement is not within noise.
[Sec. 5, FADacc definition and interpretation] The claim that CAM 'does not demonstrate degradation when generating longer sequences' is not fully supported by FADacc because the metric compares two conditions that differ in context length, not only in accumulated error. The first 10 seconds are generated from the SOS token with no prior context, while the second 10 seconds are generated after a full 10 seconds of context. The paper itself acknowledges this confound (Fig. 2b discussion: 'the Backbone receives a maximum context of ~10 seconds, it generates all embeddings after the 10 seconds mark using a full context, which may result in higher quality embeddings'). To support the error-accumulation claim, the authors should provide a controlled comparison where context length is matched, for example by measuring FAD on independently generated 10-second segments conditioned on a fixed context, or by reporting per-segment quality as a function of generation length with equivalent conditioning.
[Sec. 5, Evaluation and baselines] The comparison to the non-autoregressive baseline is limited to a single Rectified Flow model, and several experimental choices are underspecified: the number of denoising steps is selected per model from the range (10,100) 'that results in the lowest FAD' on the evaluation set, which again risks selection bias; no standard deviations or significance tests are reported; and the internal dataset is not available for independent verification. The authors should report confidence intervals, perform multiple training seeds for at least CAM and the closest baseline, and clarify whether the reported FAD numbers are best-of-N selections or averages over all evaluation draws. These additions are necessary to assess whether the improvement over MAR RF is statistically meaningful.

minor comments (5)

[Sec. 3.2 and Eq. (5)] The notation for noise level is inconsistent: Eq. (2) uses t as the time variable in Rectified Flow, while Eq. (5) reuses σ_t for the noise level, and the sampler noise level is described as sampled from a lognormal distribution. This makes it hard to distinguish the RF time variable from the noise-level variable. Please use distinct symbols (e.g., s for RF time, σ for noise level).
[Fig. 2] Figure 2 is a single panel with three subfigures (a), (b), (c), but the caption does not describe each subfigure; the text refers to 'Tab. 2c' and 'Fig. 2b' inconsistently. Please separate the tables/figures or provide a clear caption that explains what each panel shows.
[Sec. 5, Datasets] The dataset is described only as an internal collection of ~20,000 single-instrument recordings. For reproducibility, the authors should provide more details about the music styles, the instrument distribution, and whether the dataset or any subset is publicly available. The Music2Latent autoencoder is also in-house; please specify the version and any relevant training details.
[Sec. 5, Baselines] The baseline configuration details are incomplete: the exact number of denoising steps used for each diffusion-based model, the temperature and variance scaling for GIVT, and the training setup for 'GIVT+noise' are not fully specified. Please include a table with the final hyperparameter choices for every model.
[General] There are several typographical and formatting issues, including inconsistent capitalization of 'VQ-VAE', missing spaces in citations (e.g., 'VQ-V AEs' in the Introduction), and a duplicated period in the References ([4] 'Tschannen, Cian Eastwood...'). A careful proofread would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; central claim is an empirical result with externally evaluated FAD metrics.

full rationale

The paper's central claim is an empirical comparison, not a derivation: CAM is trained with Eq. (5) noise augmentation and evaluated with FAD/FADacc on generated audio against baselines. No reported quantity is fitted from the same data and then renamed a prediction. The k_inf = 0.02 value is a validation-set hyperparameter selected by optimizing FAD/FADacc (Sec. 5, Fig. 2b); reporting the resulting metrics is standard tuning, not circular, because the evaluated task (autoregressive generation of a second 10-s segment) is not identical to the tuning signal. Eq. (4) is an explicit modeling assumption about inference errors, acknowledged as such ('we assume that at inference, the Sampler generates embeddings that can be expressed as a linear combination...'), and the authors even flag its possible mismatch as a hypothesis ('We hypothesize that this noise helps to reduce the mismatch between the Gaussian distribution used for perturbation during training and the actual distribution of errors'). An unverified assumption is a correctness/robustness risk, not circularity. The only self-citation, the in-house Music2Latent encoder, is used as a preprocessing tool for all models equally and is not load-bearing for the proposed method's logical claim. Thus no circular step satisfies the quoted-reduction standard.

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

No new physical or architectural entities are introduced. The central claim rests on the assumed error model in Eq. (4), the choice of k_inf tuned to the evaluation metric, and the use of CLAP-FAD as a quality proxy. These are design choices and assumptions rather than independently validated facts.

free parameters (3)

k_inf (inference noise level) = 0.02
Chosen by evaluating FAD and FADacc across [0, 0.05] on the same benchmark used for final results (Sec. 5, Fig. 2b), so it is tuned to the evaluation metric rather than fixed a priori.
k_t sampling distribution = U(0,1)
Training-time noise augmentation level is drawn uniformly per timestep; this choice is not derived from measured inference error levels.
denoising steps per baseline = chosen from range 10-100 to minimize FAD
The paper selects the number of diffusion sampling steps that gives the lowest FAD for each model, a form of evaluation-set model selection that can favor the reported results.

assumptions (3)

ad hoc to paper Inference-time prediction errors follow x̃_t = k_t ε + (1 - k_t) x_t with ε ~ N(0,I) and k_t ~ U(0,1).
Eq. (4) in Sec. 4 is central to the training procedure. The paper does not validate this model against measured error distributions at inference.
domain assumption CLAP-based FAD is a faithful proxy for perceived audio quality.
The paper relies on a prior correlation study [23] and uses FAD as the sole quality metric, with no human listening test.
ad hoc to paper A Backbone trained without knowledge of k_t learns to distinguish real from erroneous inputs.
Sec. 4 states this as an expectation ('This results in the backbone being trained as a discriminative model'), but it is not directly demonstrated.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Continuous Autoregressive Models with Noise Augmentation Avoid Error Accumulation." pith.science (2026). https://pith.science/paper/E4HUEHIC

@misc{pith2026241118447,
  author       = {Pith},
  title        = {Pith review of: Continuous Autoregressive Models with Noise Augmentation Avoid Error Accumulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E4HUEHIC}},
  note         = {Machine review of arXiv:2411.18447}
}

read the original abstract

Autoregressive models are typically applied to sequences of discrete tokens, but recent research indicates that generating sequences of continuous embeddings in an autoregressive manner is also feasible. However, such Continuous Autoregressive Models (CAMs) can suffer from a decline in generation quality over extended sequences due to error accumulation during inference. We introduce a novel method to address this issue by injecting random noise into the input embeddings during training. This procedure makes the model robust against varying error levels at inference. We further reduce error accumulation through an inference procedure that introduces low-level noise. Experiments on musical audio generation show that CAM substantially outperforms existing autoregressive and non-autoregressive approaches while preserving audio quality over extended sequences. This work paves the way for generating continuous embeddings in a purely autoregressive setting, opening new possibilities for real-time and interactive generative applications.

Figures

Figures reproduced from arXiv: 2411.18447 by the authors.

**Figure 1.** Training process of CAM. The causal Backbone receives as input a sequence of continuous embeddings with noise augmentation. It outputs zt, which is used by the Sampler as conditioning to denoise a noise-corrupted version of xt. Autoregressive Models (AMs) have become ubiquitous in various domains, achieving remarkable success in natural language processing tasks [1, 2]. These models operate by predicting the ne… view at source ↗

**Figure 2.** (a) Comparison between MAR trained using noise-prediction with linear schedule and MAR RF using Rectified Flow. (b) Influence of kinf on FAD and FADacc. (c) Comparison of CAM with Autoregressive and Non-Autoregressive Baselines. Influence of Rectified Flow: In 2a, we first compare MAR trained using the original noise-prediction with linear schedule diffusion framework to the same model trained using a Rectified Flow… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 11 canonical work pages

[1]

Language models are unsupervised multitask learners, 2019

Alec Radford, Jeff Wu, et al. Language models are unsupervised multitask learners, 2019

work page 2019
[2]

Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwi...

2020
[3]

Neural discrete representation learning

Aäron van den Oord, Oriol Vinyals, et al. Neural discrete representation learning. In Advances in Neural Information Processing Systems 30, December 2017

work page 2017
[4]

Finite scalar quantization: VQ-V AE made simple

Fabian Mentzer, David Minnen, Eirikur Agustsson, and Michael Tschannen. Finite scalar quantization: VQ-V AE made simple. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net, 2024

work page 2024
[5]

Autoregressive image generation without vector quantization

Tianhong Li, Yonglong Tian, He Li, Mingyang Deng, and Kaiming He. Autoregressive image generation without vector quantization. arXiv preprint arXiv:2406.11838, 2024

arXiv 2024
[6]

Givt: Generative infinite-vocabulary transformers

Michael Tschannen, Cian Eastwood, and Fabian Mentzer. Givt: Generative infinite-vocabulary transformers. arXiv preprint arXiv:2312.02116, 2023

arXiv 2023
[7]

Fast transformer decoding: One write-head is all you need

Noam Shazeer. Fast transformer decoding: One write-head is all you need. arXiv preprint arXiv:1911.02150, 2019

arXiv 1911
[8]

Attention is all you need

Ashish Vaswani, Noam Shazeer, et al. Attention is all you need. In Advances in Neural Information Processing Systems 30, December 2017

work page 2017

Show all 23 references

[9]

Improving language understanding by generative pre-training, 2018

Alec Radford and Karthik Narasimhan. Improving language understanding by generative pre-training, 2018

2018
[10]

Pixel recurrent neural networks

Aäron van den Oord, Nal Kalchbrenner, and Koray Kavukcuoglu. Pixel recurrent neural networks. In Maria-Florina Balcan and Kilian Q. Weinberger, editors, Proceedings of the 33nd International Conference on Machine Learning, ICML 2016, New York City, NY, USA, June 19-24, 2016, v...

2016
[11]

WaveNet: A generative model for raw audio

Aäron van den Oord, Sander Dieleman, et al. WaveNet: A generative model for raw audio. In The 9th ISCA Speech Synthesis Workshop, September 2016

2016
[12]

Taming transformers for high-resolution image synthesis

Patrick Esser, Robin Rombach, et al. Taming transformers for high-resolution image synthesis. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2021

2021
[13]

Maskgit: Masked generative image transformer

Huiwen Chang, Han Zhang, et al. Maskgit: Masked generative image transformer. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022, 2022

2022
[14]

Jukebox: A generative model for music

Prafulla Dhariwal, Heewoo Jun, et al. Jukebox: A generative model for music. arXiv preprint arXiv:2005.00341, 2020

2005 arXiv
[15]

Simple and Controllable Music Generation, June 2023

Jade Copet, Felix Kreuk, et al. Simple and Controllable Music Generation, June 2023. arXiv:2306.05284 [cs, eess]

2023 arXiv
[16]

Flow straight and fast: Learning to generate and transfer data with rectified flow

Xingchao Liu, Chengyue Gong, and Qiang Liu. Flow straight and fast: Learning to generate and transfer data with rectified flow. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net, 2023

2023
[17]

Scaling rectified flow transformers for high-resolution image synthesis

Patrick Esser, Sumith Kulal, et al. Scaling rectified flow transformers for high-resolution image synthesis. arXiv preprint arXiv:2403.03206, 2024

2024 arXiv
[18]

Music2latent: Consistency autoencoders for latent audio compression

Marco Pasini, Stefan Lattner, and George Fazekas. Music2latent: Consistency autoencoders for latent audio compression. arXiv preprint arXiv:2408.06500, 2024

2024 arXiv
[19]

Scalable diffusion models with transformers

William Peebles and Saining Xie. Scalable diffusion models with transformers. In IEEE/CVF International Conference on Computer Vision, ICCV 2023, Paris, France, October 1-6, 2023, 2023. 6

2023
[20]

Decoupled weight decay regularization

Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenReview.net, 2019

2019
[21]

Fréchet audio distance: A reference-free metric for evaluating music enhancement algorithms

Kevin Kilgour, Mauricio Zuluaga, et al. Fréchet audio distance: A reference-free metric for evaluating music enhancement algorithms. In 20th Annual Conference of the International Speech Communication Association (INTERSPEECH), September 2019

2019
[22]

Large-scale contrastive language-audio pretraining with feature fusion and keyword-to-caption augmentation

Yusong Wu, Ke Chen, et al. Large-scale contrastive language-audio pretraining with feature fusion and keyword-to-caption augmentation. In IEEE International Conference on Acoustics, Speech and Signal Processing ICASSP 2023, Rhodes Island, Greece, June 4-10, 2023, 2023

2023
[23]

Correlation of fr \’echet audio distance with human perception of environmental audio is embedding dependant

Modan Tailleur, Junwon Lee, et al. Correlation of fr \’echet audio distance with human perception of environmental audio is embedding dependant. arXiv preprint arXiv:2403.17508, 2024. 7

2024 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Language models are unsupervised multitask learners, 2019

Alec Radford, Jeff Wu, et al. Language models are unsupervised multitask learners, 2019

work page 2019

[2] [2]

Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwi...

2020

[3] [3]

Neural discrete representation learning

Aäron van den Oord, Oriol Vinyals, et al. Neural discrete representation learning. In Advances in Neural Information Processing Systems 30, December 2017

work page 2017

[4] [4]

Finite scalar quantization: VQ-V AE made simple

Fabian Mentzer, David Minnen, Eirikur Agustsson, and Michael Tschannen. Finite scalar quantization: VQ-V AE made simple. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net, 2024

work page 2024

[5] [5]

Autoregressive image generation without vector quantization

Tianhong Li, Yonglong Tian, He Li, Mingyang Deng, and Kaiming He. Autoregressive image generation without vector quantization. arXiv preprint arXiv:2406.11838, 2024

arXiv 2024

[6] [6]

Givt: Generative infinite-vocabulary transformers

Michael Tschannen, Cian Eastwood, and Fabian Mentzer. Givt: Generative infinite-vocabulary transformers. arXiv preprint arXiv:2312.02116, 2023

arXiv 2023

[7] [7]

Fast transformer decoding: One write-head is all you need

Noam Shazeer. Fast transformer decoding: One write-head is all you need. arXiv preprint arXiv:1911.02150, 2019

arXiv 1911

[8] [8]

Attention is all you need

Ashish Vaswani, Noam Shazeer, et al. Attention is all you need. In Advances in Neural Information Processing Systems 30, December 2017

work page 2017