Pith. sign in

REVIEW 2 major objections 5 minor 2 cited by

FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow

T0 review · 2 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read FlowSeq uses a generative-flow prior to decode all target tokens in a single parallel pass.

desk verdict FlowSeq is a genuine advance for non-autoregressive NMT, but its 'almost constant decoding time' claim is not backed by the architecture or the experiments. read the letter →

arxiv 1909.02480 v3 pith:526XUK4W submitted 2019-09-05 cs.CL cs.LG

Xuezhe Ma , Chunting Zhou , Xian Li , Graham Neubig , Eduard Hovy This is my paper · ORCID
classification cs.CLcs.LG
keywords FlowSeqnon-autoregressivegenerationneuralmachinetranslationnormalizingflowslatentvariablemodelsparalleldecodinggenerativevariationalinference
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

FlowSeq proposes a non-autoregressive sequence-to-sequence model in which the target sentence is generated from a latent sequence of continuous vectors, sampled from a normalizing-flow prior conditioned on the source. Because the decoder factorizes over target positions given the latent sequence, all output tokens are produced in one parallel pass rather than left to right. The paper argues that the flow prior is expressive enough to absorb the interdependencies between target words that a naive independent-token model misses. On WMT14, WMT16, and IWSLT14 translation benchmarks, FlowSeq reports translation quality comparable to state-of-the-art non-autoregressive systems, with decoding time nearly constant as target length grows. If the claim holds, the usual accuracy-parallelism trade-off in non-autoregressive generation is substantially looser than previously demonstrated.

What carries the argument

The load-bearing mechanism is the flow-based conditional prior over the latent sequence $z$. The prior is a chain of invertible transformations: actnorm layers normalize activations; multi-head invertible linear layers mix feature dimensions at $O(d_h^3)$ cost per head rather than $O(d_z^3)$; affine coupling layers, with split patterns over time and feature dimensions, propagate source information through attention and model interdependence across time steps; and a multi-scale architecture halves dimensions after each scale. The invertibility and tractable Jacobian determinants let the model compute the density $p_{\theta}(z|x)$ exactly during training, while the inverse map lets it sample $z$ from the prior at test time. This prior is what carries the argument: it is asked to encode enough structure about the target sentence that the conditionally independent decoder can read off a coherent translation in one pass.

What would settle it

On WMT2014 DE-EN, retrain FlowSeq with the multi-step flow prior replaced by a diagonal-Gaussian prior of the same latent dimension, keeping decoder, posterior, loss, and decoding fixed; if argmax BLEU does not drop substantially, the flow layers are not carrying the translation quality the paper assigns to them.

Watch

Extended reading notes

Core claim

The central claim is that generative flows can serve as the prior $p_{\theta}(z|x)$ in a latent-variable non-autoregressive seq2seq model, making the factorized decoder $P_{\theta}(y|z,x)=\prod_t P_{\theta}(y_t|z,x)$ accurate enough for competitive machine translation. FlowSeq models $z$ as a length-$T$ sequence of continuous vectors, trains the posterior $q_{\varphi}(z|y,x)$ as a diagonal Gaussian, and trains the prior as a deep invertible flow composed of actnorm, multi-head invertible linear layers, and source-conditioned affine coupling layers in a multi-scale architecture. At test time, $z$ is sampled from the prior and all target tokens are decoded in parallel; candidates can be reranked by noisy parallel decoding or importance-weighted decoding. The paper reports that FlowSeq with pure argmax decoding improves over earlier non-autoregressive baselines by more than 9 BLEU points on raw data, and with rescoring reaches performance close to contemporaneous state-of-the-art non-autoregressive models. It also reports that decoding time stays almost constant as sentence length increases, in contrast to the linear growth of an autoregressive Transformer.

Load-bearing premise

The load-bearing premise is that the trained flow prior packs enough of the target sentence's structure into the latent sequence that the decoder, which reads each word independently given that sequence, still produces high-quality translations; the paper protects this premise with an unprincipled token-dropout heuristic rather than a representational analysis.

Editorial extensions

If this is right

  • Non-autoregressive translation can reach competitive quality without iterative refinement, at least when rescoring with NPD or IWD is allowed.
  • The benefit of knowledge distillation from an autoregressive teacher is smaller for FlowSeq than for earlier non-autoregressive models, with no reported improvement on WMT2016 RO-EN.
  • Because FlowSeq's decoding time stays nearly constant as target sentences grow, the speed advantage over autoregressive decoding widens with sequence length.
  • The parallel, unmasked decoder is compatible with iterative refinement, a combination the paper leaves to future work.

Reading between the lines

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

  • A natural ablation is to replace token dropout with a posterior regularizer that penalizes the mutual information between each latent position $z_t$ and the corresponding target token $y_t$; if BLEU holds, the heuristic is not necessary.
  • The continuous, invertible latent space invites controlled-generation experiments, such as interpolating between latent codes of two source sentences to steer output style or content, which the paper does not test.
  • The length predictor's range of $[-20, 20]$ limits FlowSeq to near-length-preserving outputs; tasks with large length mismatches would need a different length model before the flow prior can show its value.
  • The paper's own sample outputs show repetition and broken translations that it attributes to weak direct modeling of target-word dependencies; quantifying how these errors scale with length and sampling temperature would sharpen the case for the flow prior.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. FlowSeq proposes a non-autoregressive sequence-to-sequence model that uses generative flows to model a latent sequence z, from which the target tokens are decoded independently given the source x. Training maximizes an ELBO with a Gaussian posterior q_phi(z|y,x) and a flow-based prior p_theta(z|x); at test time, z is sampled from the prior and the target is predicted in a single parallel pass. The paper reports BLEU scores on WMT14, WMT16, and IWSLT14, compares against several non-autoregressive baselines, and measures decoding speed relative to a Transformer. The central claims are that FlowSeq achieves performance comparable to state-of-the-art non-autoregressive models and that decoding time is almost constant with respect to sequence length.

Significance. If the results hold, FlowSeq is a meaningful contribution to non-autoregressive sequence generation. It provides a principled latent-variable framework based on normalizing flows, with a multi-scale architecture and invertible layers, and it is among the first to apply flows to seq2seq. The paper is clearly written, the mathematical framework (ELBO, change-of-variables) is sound, and the code is released. The empirical results show that FlowSeq is competitive with strong non-autoregressive baselines, especially in the raw-data setting, and the diversity analysis is a useful addition. The main reservation is that the headline efficiency claim is not supported by the architecture or the measured length range, which affects the paper's central contribution.

major comments (2)
  1. [Abstract; Section 4.3, Figure 4b] The claim of 'almost constant decoding time w.r.t. the sequence length' is not substantiated and is in tension with the architecture. The prior flow has 112 steps (Section 3.4), and each coupling layer's s(za,x) and b(za,x) are implemented with a Transformer decoder layer performing unmasked multi-head self-attention over the time dimension. The decoder (Section 3.3) also uses unmasked self-attention. The asymptotic cost per sequence is therefore at least O(T^2), so per-sentence decoding time should grow with T once attention dominates fixed overhead. Figure 4b only reports lengths up to the '>40' bucket (with training capped at length 80), so the flat curve is consistent with a fixed-overhead regime rather than true length-invariance. The authors should either provide measurements for longer sequences (e.g., up to the model's maximum length and beyond) or qualify the claim to a specific length regime. As written, the efficiency half of the central claim is unsupported.
  2. [Section 3.2 (Token Dropout); Section 4.2] The paper posits that the expressiveness of the flow prior is the key to making the factorized decoder (Eq. 4) work, but the only safeguard against posterior collapse to the naive independent-token model is the heuristic token dropout. No ablation is reported on the dropout probability, nor is any analysis provided of the information content of the latent codes (e.g., probing or visualization). Because the entire modeling bet rests on this mechanism, the paper would be stronger with an ablation study or a diagnostic showing that z captures target-side dependencies. Without this, the attribution of the performance gain to the flow prior rather than to the training heuristic remains an unverified claim.
minor comments (5)
  1. [Section 4.2] The statement that knowledge distillation yields 'even no improvement on WMT2016 RO-EN corpus' is contradicted by Table 1, which shows an improvement from 30.16 to 30.44 BLEU for FlowSeq-base on RO-EN.
  2. [Appendix B, Table 3] Table 3 contains typographical errors: 'FlowSeq=large' should be 'FlowSeq-large', and the dimensions '512/2014' and '2014/4096' should likely be '512/1024' and '1024/4096', respectively.
  3. [Section 4.3, Figure 4] The decoding speed plots do not include error bars or multiple-run variance, so it is unclear whether the differences across length buckets are significant; please report standard deviations or confidence intervals.
  4. [Section 3.2] The probability of token dropout is not specified in the hyperparameter description; please provide the value or range used and report sensitivity to this hyperparameter.
  5. [Section 4.4, Figure 5] The text says the sampling temperature was selected from {0.1, 0.2, 0.3, 0.4, 0.5, 1.0}, but the x-axis of Figure 5 begins at 0.2; consider aligning the axis with the stated range.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: FlowSeq's claims are empirical results from standard train/dev/test evaluation, and no conclusion reduces by construction to a fitted input or to a self-citation.

full rationale

The paper's central claims are empirical, not derived from the model's own inputs in a circular way. The model is defined by the ELBO objective in Eq. (8) over the latent-variable factorization in Eqs. (3)-(4), and translation quality is measured with held-out BLEU on standard benchmarks (WMT2014, WMT2016, IWSLT2014). No parameter is fit to the test set and then reported as a prediction: hyperparameters are selected on development sets, as stated in Section 4.2, and the temperature is selected based on dev performance. The decoding-time claim in Section 4.3 and Figure 4b is an empirical measurement; even if the 'almost constant' claim is in tension with the unmasked self-attention inside the flow coupling layers, that is a potential support or correctness issue, not circularity. The paper's self-citations (Ma et al. 2018, 2019; Ma and Hovy 2019) are used only for peripheral context: token dropout similarity, KL-collapse observations, and prior flow-based modeling success. They do not carry the load of the BLEU or speed results. The advanced decoding procedures NPD and IWD use autoregressive rescoring or importance weighting; these are external scoring mechanisms, not equations that assume the target result. The Appendix explicitly acknowledges a limitation: 'repetition and broken translations also exist in some cases due to the lack of direct modeling of dependencies between target words.' That is an honest limitation, not a circular step. Overall, this is a self-contained empirical paper with no circular derivation chain.

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

The model's central performance depends on several hand-chosen training schedules and a restricted length range. The architectural assumptions (factorized decoder, diagonal Gaussian posterior) are standard for non-autoregressive latent-variable models, but they are not derived from first principles. No new physical or conceptual entities are introduced.

free parameters (4)
  • KL weight schedule = 0 for first 30k updates, linear to 1 over next 10k
    Described as essential for stable training and performance (Section 4.1); no principled derivation.
  • Token dropout probability = not stated in the paper (implementation detail)
    Token dropout in the posterior is introduced to avoid the local optimum where z_t encodes only y_t; the exact probability is not given, making the method sensitive to an unspecified hyperparameter.
  • Sampling temperature tau = selected from {0.1, 0.2, 0.3, 0.4, 0.5, 1.0} on dev set
    Used for IWD/NPD sampling; the best value varies by setting and is chosen by evaluating BLEU on dev, i.e., a fitted hyperparameter.
  • Length-difference classifier range = [-20, 20]
    The model can only represent target lengths within 20 tokens of the source length; this limits applicability to sentences with larger length ratios.
assumptions (5)
  • standard math Change-of-variable formula for invertible transformations
    Used in Eq. (6) to compute the flow density; standard result from calculus.
  • standard math Variational inference ELBO
    Used in Eq. (8) to train the latent variable model; standard lower bound.
  • ad hoc to paper The decoder factorizes independently given z (Eq. 4)
    This is the non-autoregressive assumption; it is a modeling choice that offloads all dependency modeling onto z and the flow prior, and if it is too strong the model degrades to Eq. (2).
  • ad hoc to paper Posterior q_phi(z|y,x) is a product of diagonal Gaussians (Eq. 9)
    A simplifying assumption for tractability; no evidence that the true posterior is Gaussian or that the diagonal approximation is sufficient.
  • domain assumption Target length difference is within [-20,20]
    The length predictor in Section 3.5 can only output differences in this range; sentences with larger length differences cannot be generated with the correct length.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow." pith.science (2026). https://pith.science/paper/526XUK4W

@misc{pith2026190902480,
  author       = {Pith},
  title        = {Pith review of: FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/526XUK4W}},
  note         = {Machine review of arXiv:1909.02480}
}
read the original abstract

Most sequence-to-sequence (seq2seq) models are autoregressive; they generate each token by conditioning on previously generated tokens. In contrast, non-autoregressive seq2seq models generate all tokens in one pass, which leads to increased efficiency through parallel processing on hardware such as GPUs. However, directly modeling the joint distribution of all tokens simultaneously is challenging, and even with increasingly complex model structures accuracy lags significantly behind autoregressive models. In this paper, we propose a simple, efficient, and effective model for non-autoregressive sequence generation using latent variable models. Specifically, we turn to generative flow, an elegant technique to model complex distributions using neural networks, and design several layers of flow tailored for modeling the conditional density of sequential latent variables. We evaluate this model on three neural machine translation (NMT) benchmark datasets, achieving comparable performance with state-of-the-art non-autoregressive NMT models and almost constant decoding time w.r.t the sequence length.

Figures

Figures reproduced from arXiv: 1909.02480 by the authors.

Figure 1
Figure 1. (a) Autoregressive (b) non-autoregressive and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Neural architecture of FlowSeq, including the encoder, the decoder and the posterior networks, together [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. (a) The architecture of one step of our flow. (b) The visualization of three split pattern for coupling [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: The decoding speed of the Transformer (batched, beam size 5) and FlowSeq on WMT14 EN-DE test set [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Impact of sampling hyperparameters on the [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 7
Figure 7. Figure 7: Training dynamics. In [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. PhyloVAE: Unsupervised Learning of Phylogenetic Trees via Variational Autoencoders

    stat.ML 2025-02 conditional novelty 7.0 of 10

    PhyloVAE is a deep variational autoencoder that learns low-dimensional latent representations of tree topologies and generates new trees non-autoregressively, faster than ARTree.

  2. LTM3D: Bridging Token Spaces for Conditional 3D Generation with Auto-Regressive Diffusion Framework

    cs.CV 2025-05 conditional novelty 5.0 of 10

    A conditional 3D generation framework that combines masked autoencoding and diffusion in token space, with prefix learning and reconstruction-guided sampling, reports state-of-the-art results on ShapeNet and Objaverse.

Reference graph

Works this paper leans on

37 extracted references · 22 canonical work pages · cited by 2 Pith papers

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2015. Neural machine translation by jointly learning to align and translate. In International Conference on Learning Representations (ICLR)

    2015

  4. [4]

    Samuel R Bowman, Luke Vilnis, Oriol Vinyals, Andrew M Dai, Rafal Jozefowicz, and Samy Bengio. 2015. Generating sentences from a continuous space. arXiv preprint arXiv:1511.06349

    arXiv 2015

  5. [5]

    Yuri Burda, Roger Grosse, and Ruslan Salakhutdinov. 2015. Importance weighted autoencoders. arXiv preprint arXiv:1509.00519

    arXiv 2015

  6. [6]

    Mauro Cettolo, Christian Girardi, and Marcello Federico. 2012. Wit3: Web inventory of transcribed and translated talks. In Conference of European Association for Machine Translation, pages 261--268

    work page 2012

  7. [7]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805

    arXiv 2018

  8. [8]

    Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio. 2016. Density estimation using real nvp. arXiv preprint arXiv:1605.08803

    arXiv 2016

Show all 37 references
  1. [9]

    Marjan Ghazvininejad, Omer Levy, Yinhan Liu, and Luke Zettlemoyer. 2019. Constant-time machine translation with conditional masked language models. arXiv preprint arXiv:1904.09324

    2019 arXiv

  2. [10]

    Jiatao Gu, James Bradbury, Caiming Xiong, Victor OK Li, and Richard Socher. 2018. Non-autoregressive neural machine translation. Proceedings of the 6th International Conference on Learning Representations (ICLR-2018)

    2018

  3. [11]

    Jiatao Gu, Qi Liu, and Kyunghyun Cho. 2019. Insertion-based decoding with automatically inferred generation order. arXiv preprint arXiv:1902.01370

    2019 arXiv

  4. [12]

    Xuanli He, Gholamreza Haffari, and Mohammad Norouzi. 2018. Sequence to sequence mixture model for diverse machine translation. In Proceedings of the 22nd Conference on Computational Natural Language Learning, pages 583--592

    2018

  5. [13]

    Sergey Ioffe and Christian Szegedy. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In International Conference on Machine Learning, pages 448--456

    2015

  6. [14]

    Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980

    2014 arXiv

  7. [15]

    Diederik P Kingma, Tim Salimans, Rafal Jozefowicz, Xi Chen, Ilya Sutskever, and Max Welling. 2016. Improving variational inference with inverse autoregressive flow. The 29th Conference on Neural Information Processing Systems

    2016

  8. [16]

    Durk P Kingma and Prafulla Dhariwal. 2018. Glow: Generative flow with invertible 1x1 convolutions. In Advances in Neural Information Processing Systems, pages 10215--10224

    2018

  9. [17]

    Jason Lee, Elman Mansimov, and Kyunghyun Cho. 2018. Deterministic non-autoregressive neural sequence modeling by iterative refinement. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 1173--1182

    2018

  10. [18]

    Jind r ich Libovick \`y and Jind r ich Helcl. 2018. End-to-end non-autoregressive neural machine translation with connectionist temporal classification. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 3016--3021

    2018

  11. [19]

    Xuezhe Ma and Eduard Hovy. 2019. Macow: Masked convolutional generative flow. arXiv preprint arXiv:1902.04208

    2019 arXiv

  12. [20]

    Xuezhe Ma, Zecong Hu, Jingzhou Liu, Nanyun Peng, Graham Neubig, and Eduard Hovy. 2018. Stack-pointer networks for dependency parsing. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1403--1414

    2018

  13. [21]

    Xuezhe Ma, Chunting Zhou, and Eduard Hovy. 2019. Mae: Mutual posterior-divergence regularization for variational autoencoders. In Proceedings of the 7th International Conference on Learning Representations (ICLR-2019), New Orleans, Louisiana, USA

    2019

  14. [22]

    Oren Melamud, Jacob Goldberger, and Ido Dagan. 2016. https://doi.org/10.18653/v1/K16-1006 context2vec: Learning generic context embedding with bidirectional LSTM . In Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning , pages 51--61, Berlin, G...

    2016 doi

  15. [23]

    Myle Ott, Michael Auli, David Grangier, et al. 2018. Analyzing uncertainty in neural machine translation. In International Conference on Machine Learning, pages 3953--3962

    2018

  16. [24]

    Myle Ott, Sergey Edunov, Alexei Baevski, Angela Fan, Sam Gross, Nathan Ng, David Grangier, and Michael Auli. 2019. fairseq: A fast, extensible toolkit for sequence modeling. In Proceedings of NAACL-HLT 2019: Demonstrations

    2019

  17. [25]

    George Papamakarios, Theo Pavlakou, and Iain Murray. 2017. Masked autoregressive flow for density estimation. In Advances in Neural Information Processing Systems, pages 2338--2347

    2017

  18. [26]

    Ryan Prenger, Rafael Valle, and Bryan Catanzaro. 2019. Waveglow: A flow-based generative network for speech synthesis. In ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 3617--3621. IEEE

    2019

  19. [27]

    Sashank J Reddi, Satyen Kale, and Sanjiv Kumar. 2018. On the convergence of adam and beyond. In Proceedings of the 6th International Conference on Learning Representations (ICLR-2018)

    2018

  20. [28]

    Danilo Jimenez Rezende and Shakir Mohamed. 2015. Variational inference with normalizing flows. In Proceedings of the 32nd International Conference on International Conference on Machine Learning-Volume 37, pages 1530--1538. JMLR. org

    2015

  21. [29]

    Alexander M Rush, Sumit Chopra, and Jason Weston. 2015. A neural attention model for abstractive sentence summarization. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 379--389

    2015

  22. [30]

    Tianxiao Shen, Myle Ott, Michael Auli, et al. 2019. Mixture models for diverse machine translation: Tricks of the trade. In International Conference on Machine Learning, pages 5719--5728

    2019

  23. [31]

    Raphael Shu, Jason Lee, Hideki Nakayama, and Kyunghyun Cho. 2019. Latent-variable non-autoregressive neural machine translation with deterministic inference using a delta posterior. arXiv preprint arXiv:1908.07181

    2019 arXiv

  24. [32]

    Mitchell Stern, William Chan, Jamie Kiros, and Jakob Uszkoreit. 2019. Insertion transformer: Flexible sequence generation via insertion operations. arXiv preprint arXiv:1902.03249

    2019 arXiv

  25. [33]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in neural information processing systems, pages 5998--6008

    2017

  26. [34]

    Oriol Vinyals, Alexander Toshev, Samy Bengio, and Dumitru Erhan. 2015. Show and tell: A neural image caption generator. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3156--3164

    2015

  27. [35]

    Martin J Wainwright, Michael I Jordan, et al. 2008. Graphical models, exponential families, and variational inference. Foundations and Trends in Machine Learning , 1(1--2):1--305

    2008

  28. [36]

    Yiren Wang, Fei Tian, Di He, Tao Qin, ChengXiang Zhai, and Tie-Yan Liu. 2019. Non-autoregressive machine translation with auxiliary regularization. arXiv preprint arXiv:1902.10245

    2019 arXiv

  29. [37]

    Zachary Ziegler and Alexander Rush. 2019. Latent normalizing flows for discrete sequences. In International Conference on Machine Learning, pages 7673--7682

    2019

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.