Pith. sign in

REVIEW 3 major objections 5 minor 46 references

A Decoding Algorithm for Length-Control Summarization Based on Directed Acyclic Transformers

T0 review · 3 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read This paper proposes SeqMAP, a decoding objective for Directed Acyclic Transformers that marginalizes over all link paths and selects the most probable summary of a fixed length, reporting higher ROUGE than PathMAP and CTC baselines with…

desk verdict SeqMAP is a plausible and well-tested decoding improvement over PathMAP for DAT length control, but the paper overstates the training-alignment story and tunes hyperparameters on the test set. read the letter →

arxiv 2502.04535 v1 pith:S5BDMUJZ submitted 2025-02-06 cs.CL

Chenyang Huang , Hao Zhou , Cameron Jen , Kangjie Zheng , Osmar R. Zaïane , Lili Mou This is my paper · ORCID
classification cs.CL
keywords length-controlsummarizationdirectedacyclictransformernon-autoregressivegenerationSeqMAPdecodingbeamsearchdynamicprogrammingsequencererankinghardlengthconstraintROUGEevaluation
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

Summaries that must fit a fixed length—web titles, tweets, abstract limits—are hard to generate because most models treat length as a soft constraint and then truncate. This paper adapts the Directed Acyclic Transformer, a non-autoregressive model that writes a summary as a graph of word fragments and links, and proposes SeqMAP: rather than committing to the single most probable path through the graph, it sums the probabilities of every path that spells the same candidate summary and then picks the highest-scoring summary of the target length. That sum cannot be maximized by standard dynamic programming, so the authors develop an approximate beam-search algorithm with expand-and-merge recursion and a learned reranker over the beam. On Gigaword and DUC2004, SeqMAP achieves exact length control with zero truncated outputs and reports higher ROUGE scores than CTC-based length control and the earlier PathMAP decoding.

What carries the argument

The central object is the Directed Acyclic Transformer (DAT), a non-autoregressive model that maintains $S$ prediction steps, each with a word distribution and a link distribution over future steps; a path is a chain of linked steps with predicted words. The SeqMAP objective (Eq. 11) is a sum over every valid path and ending step of the product of link and word probabilities for a given word sequence. The algorithm that carries the argument is an approximate beam-search dynamic program with two operations: EXPAND, which extends length-$(t-1)$ beam sequences with the top-$V$ words at step $s$ and scores each extension by marginalizing over the previous linked step's probability, and MERGE, which keeps the top-$K$ sequences generated at or before step $s$. A reranker built on a pretrained RoBERTa encoder with rank embeddings then picks the best candidate in the final beam.

What would settle it

Take a small synthetic DAT with known word and link probabilities, with vocabulary and steps small enough to enumerate all length-$T$ word sequences and all link paths; compute the exact SeqMAP argmax from Eq. (11) and compare it with the output of the beam-search algorithm. If the approximate algorithm frequently returns a sequence whose SeqMAP score is below the exact optimum, the empirical gains could stem from the search heuristic rather than from the marginalization principle.

Watch

Extended reading notes

Core claim

The central claim is that the best fixed-length summary under a Directed Acyclic Transformer (DAT) is found not by taking the single most probable path of linked words (the PathMAP objective) but by a Sequence Maximum a Posteriori (SeqMAP) objective: for a candidate word sequence of length $T$, sum the probabilities of all link paths and all valid ending steps that spell that sequence, then choose the sequence with the largest sum. Because DAT's training objective marginalizes over link paths, SeqMAP aligns decoding with training better than PathMAP. The paper develops an approximate beam-search dynamic program for this objective—exact search is intractable because the summation and maximization cannot be swapped—and shows empirically that SeqMAP, with and without a learned reranker, outperforms PathMAP and CTC-based length-control summarization on Gigaword and DUC2004 while producing zero truncated outputs.

Load-bearing premise

The load-bearing premise is that the beam-search approximation—keeping only the top-$K$ partial sequences and top-$V$ word expansions at each step—still contains the true argmax of the SeqMAP objective, and that the objective itself matches DAT's training marginalization; if either fails, the measured gains could come from the search heuristic rather than from the marginalization principle.

Editorial extensions

If this is right

  • Exact length control becomes practical for headline, tweet, and abstract generation: every output meets the mandated length, eliminating the truncation step that soft-constraint models require.
  • Marginalizing over link paths rather than committing to one path improves ROUGE-Sum by roughly 1 to 2.5 points over PathMAP and by 6 to 8 points over CTC on the two reported datasets.
  • The alignment between decoding and DAT's marginalization-based training suggests the SeqMAP principle should transfer to other DAT-based generation tasks beyond summarization.
  • The reranker adds about 0.5 to 1.3 ROUGE-Sum points on top of SeqMAP, and its rank embeddings produce a consistent gain in the ablation, indicating that the model's own ranking is useful signal.
  • Non-autoregressive decoding with SeqMAP (without the reranker) is faster than an autoregressive Transformer baseline, and even the reranked version remains competitive in speed while enforcing the length budget.

Reading between the lines

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

  • The same SeqMAP-style sum-over-paths decoding could apply to any non-autoregressive generator whose output space is a DAG or lattice, such as CTC-based models, by replacing a Viterbi-style path choice with sequence-level marginalization; this is not tested in the paper.
  • Because SeqMAP sums over paths ending at any valid step while DAT training marginalizes over paths ending at the final step $S$, a natural testable extension is to train DAT with a length-agnostic marginalization over all end steps, which might align training and inference even more closely.
  • The gap between the approximate beam search and the exact SeqMAP argmax can be measured on small synthetic instances (small $S$, small vocabulary, known probabilities) by brute-force enumeration; if the beam frequently misses the true optimum, the reported gains are at least partly due to search, not marginalization alone.
  • The rank-embedding reranker design is portable to other n-best reranking problems where the base model's ranking carries information; the paper's ablation shows this component alone is worth about 0.6 ROUGE-Sum.
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

3 major / 5 minor

Summary. The paper adapts the Directed Acyclic Transformer (DAT) to length-control summarization and proposes a Sequence Maximum a Posteriori (SeqMAP) decoding objective that marginalizes over possible linked steps, together with an approximate beam-search dynamic programming algorithm and a BERT/RoBERTa-based reranker. Experiments on Gigaword and DUC2004 report consistent ROUGE improvements over PathMAP and CTC baselines, with exact length control, and the paper releases code and outputs.

Significance. If the central claims hold, the paper makes a useful contribution: it brings a more expressive non-autoregressive model to length-controlled summarization, gives a concrete decoding algorithm, and shows consistent empirical gains. The availability of code and outputs is a strength. However, the theoretical motivation—that SeqMAP aligns with DAT's training objective—is not supported by the equations as written, and the empirical evidence lacks significance testing and a clearly separated validation split for hyperparameter selection. These issues bear directly on the paper's main claims and need to be resolved before the results can be fully credited.

major comments (3)
  1. [§2.2, Eq. (11); §2.1, Eqs. (3)–(7)] The SeqMAP objective in Eq. (11) is not the marginal sequence probability of the DAT model defined in §2.1. Training in Eq. (7) marginalizes only over paths a in Γ_{Ty,S} with a_1=1 and a_{Ty}=S, and, per footnote 2, every path includes <bos> at step 1 and <eos> at step S. Eq. (11), by contrast, sums over paths ending at any step s in {T,...,S} and omits both the mandatory final link l_{a_T,S} and the <eos> word factor. Thus Eq. (11) is a different probability model from the one used in training. The paper's central motivation in §1 and §2.2—that SeqMAP 'aligns with the training objective' because it marginalizes links—is therefore unsupported. Please either redefine SeqMAP as a proper marginalization over Γ_{T+2,S} with <bos>/<eos> and the final link to step S, or explicitly present Eq. (11) as a heuristic decoding objective and remove the training-inference alignment claim.
  2. [§3.2, Tables 7 and 8; §3.1 Implementation Details] The operating hyperparameters K=20 and V=5 are reported in §3.1, and their selection is supported by Tables 7 and 8, but the paper never states whether these tables report validation or test ROUGE. If the tables are on the Gigaword test set, then K and V have been tuned on the test data, which can inflate the reported gains. This is especially important because the SeqMAP-versus-PathMAP differences in R-Sum are around 1 point. Please specify the split used for hyperparameter selection, report multiple seeds or confidence intervals, and provide significance tests for the main ROUGE comparisons.
  3. [§2.2, Algorithm 1; §6 Limitations] The paper's limitations section correctly concedes that Algorithm 1 does not guarantee finding the exact SeqMAP optimum, but the empirical section interprets the SeqMAP-vs-PathMAP gap as evidence of 'the superiority of the SeqMAP objective.' Because Algorithm 1 combines the new objective with top-K/top-V pruning, the observed gains could in principle come from the search heuristic rather than from marginalization. Please provide a diagnostic that separates objective from search—for example, exact or near-exact computation on small S, or a comparison where both methods use matched search effort—so that the reader can tell which component drives the improvement.
minor comments (5)
  1. [§1] In the sentence 'our SeqMAP aligns with the training objective better (as it huggalso performs marginalization)', 'huggalso' appears to be a typo for 'also'.
  2. [§3.2, Table 8 caption] The caption says 'Reranker performance of different K' but the table varies V with K fixed at 20; the caption should say 'different V'.
  3. [§3.1] The text says 'the DUC2024 experiment' but the dataset is DUC2004; please correct the typo.
  4. [§2.1, footnote 2] The notation is ambiguous about whether y already includes <bos> and <eos>; if the special tokens are added, the equations and the SeqMAP objective should be stated consistently in terms of the augmented sequence length.
  5. [Appendix A, Algorithm 1] Line 12 contains an extraneous closing brace in 'top-K{Bt,s}' that should be removed, and the ranking criterion for B_{t,s} should be stated explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the SeqMAP objective is an explicit decoding criterion over DAT model probabilities, Algorithm 1 is an approximate optimizer, and the reranker is trained and evaluated against held-out references.

full rationale

The derivation chain is self-contained. DAT supplies word and link probabilities from a trained model; SeqMAP (Eq. 11) is an explicit decoding objective over those probabilities, and Algorithm 1 is an approximate beam-search optimizer of that objective. The reranker is trained on training-sample word overlap with groundtruth and applied to beam candidates, then assessed on Gigaword and DUC2004 test references, so the evaluation is not determined by a fitted constant or by a parameter renamed as a prediction. Self-citations to Huang et al. (2022b) for the DAT backbone and to Shao et al. (2022) for PathMAP are conventional background and do not carry the central argument. The Limitations passage concedes that the proposed algorithm does not guarantee finding the best sequence in DAT's decoding space; that is an honest search-optimality caveat rather than evidence of circularity. The main validity concern is that Eq. (11) marginalizes over paths ending at any step s and omits the final bos/eos structure used in training (Eq. 6), so the claimed training-inference alignment is not exact; however, this is a correctness or objective-matching issue, not a circular reduction where the reported results are forced by construction. No step in the paper equates its output with its own input, and the empirical claims are measured against unseen groundtruth summaries.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the trained DAT probability model, on the beam approximation to an intractable marginalization, and on the reranker's overlap-based training signal. The only hand-tuned decoding parameters explicitly varied and selected are K and V; no new entities are introduced.

free parameters (2)
  • Beam size K = 20
    Selected from ablation in Table 7; performance peaks around K=20 on the reported ROUGE curves. This is a decoding hyperparameter that the reported gains depend on.
  • Vocabulary exploration size V = 5
    Selected from ablation in Table 8; with fixed K=20, V=5 gives the best R-Sum. It controls how many words are expanded per step in the approximate DP.
assumptions (4)
  • domain assumption DAT's link and word probabilities, as trained by marginalizing over paths, are well-calibrated enough that maximizing or marginalizing them selects good summaries.
    Section 2.1 treats p_link and p_word as the scoring functions; no calibration or quality adjustment is made for decoding.
  • ad hoc to paper SeqMAP's sum over paths ending at any step s (Eq. 11) is the correct training-consistent objective, even though DAT training marginalizes over paths whose final linked step is S.
    Eqs. 8 and 11 allow s in {T,...,S}; DAT training paths always end at S per the definition after Eq. 3. This mismatch is not discussed.
  • ad hoc to paper The beam search keeps enough probability mass that top-K/top-V truncation preserves the optimal sequence.
    Algorithm 1 and Section 2.2; the Limitations section explicitly says no guarantee of finding the best sequence.
  • domain assumption Overlap with the groundtruth summary (word overlap) is a valid training signal for the reranker and a valid proxy for summary quality.
    Section 2.3 trains the reranker to predict the beam candidate with the most overlapped words with y; evaluation uses ROUGE, another overlap metric.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Decoding Algorithm for Length-Control Summarization Based on Directed Acyclic Transformers." pith.science (2026). https://pith.science/paper/S5BDMUJZ

@misc{pith2026250204535,
  author       = {Pith},
  title        = {Pith review of: A Decoding Algorithm for Length-Control Summarization Based on Directed Acyclic Transformers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S5BDMUJZ}},
  note         = {Machine review of arXiv:2502.04535}
}
read the original abstract

Length-control summarization aims to condense long texts into a short one within a certain length limit. Previous approaches often use autoregressive (AR) models and treat the length requirement as a soft constraint, which may not always be satisfied. In this study, we propose a novel length-control decoding algorithm based on the Directed Acyclic Transformer (DAT). Our approach allows for multiple plausible sequence fragments and predicts a \emph{path} to connect them. In addition, we propose a Sequence Maximum a Posteriori (SeqMAP) decoding algorithm that marginalizes different possible paths and finds the most probable summary satisfying the length budget. Our algorithm is based on beam search, which further facilitates a reranker for performance improvement. Experimental results on the Gigaword and DUC2004 datasets demonstrate our state-of-the-art performance for length-control summarization.

Figures

Figures reproduced from arXiv: 2502.04535 by the authors.

Figure 1
Figure 1. The neural architecture of our reranker. This [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 38 canonical work pages

  1. [1]

    Samy Bengio, Oriol Vinyals, Navdeep Jaitly, and Noam Shazeer. 2015. https://proceedings.neurips.cc/paper/2015/file/e995f98d56967d946471af29d7bf99f1-Paper.pdf Scheduled sampling for sequence prediction with recurrent neural networks . In Advances in Neural Information Processing Systems, pages 1171--1179

    work page 2015

  2. [2]

    Rishi Bommasani and Claire Cardie. 2020. https://aclanthology.org/2020.emnlp-main.649 Intrinsic evaluation of summarization datasets . In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 8075--8096

    work page 2020

  3. [3]

    Shun-Po Chuang, Yung-Sung Chuang, Chih-Chiang Chang, and Hung-yi Lee. 2021. https://aclanthology.org/2021.findings-acl.92 Investigating the reordering capability in CTC -based non-autoregressive end-to-end speech translation . In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 1068--1077

    work page 2021

  4. [4]

    Cassio P de Campos. 2011. https://arxiv.org/abs/1007.3884 New complexity results for map in bayesian networks . In Proceedings of the International Joint Conference on Artificial Intelligence, pages 2100--2106

    work page Pith review arXiv 2011

  5. [5]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://aclanthology.org/N19-1423 BERT : Pre-training of deep bidirectional transformers for language understanding . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and ...

    work page 2019

  6. [6]

    Chris Dyer, Victor Chahuneau, and Noah A. Smith. 2013. https://aclanthology.org/N13-1073 A simple, fast, and effective reparameterization of IBM M odel 2 . In Proceedings of the 2013 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies , pages 644--648

    2013

  7. [7]

    David Graff, Junbo Kong, Ke Chen, and Kazuaki Maeda. 2003. https://catalog.ldc.upenn.edu/LDC2003T05 English G igaword . Linguistic Data Consortium, 4:34

    work page 2003

  8. [8]

    Alex Graves, Santiago Fern \'a ndez, Faustino Gomez, and J \"u rgen Schmidhuber. 2006. https://dl.acm.org/doi/abs/10.1145/1143844.1143891 Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks . In Proceedings of the 23rd International Conference on Machine learning, pages 369--376

    work page arXiv 2006

Show all 46 references
  1. [9]

    Li, and Richard Socher

    Jiatao Gu, James Bradbury, Caiming Xiong, Victor O.K. Li, and Richard Socher. 2018. https://openreview.net/forum?id=B1l8BtlCb Non-autoregressive neural machine translation . In International Conference on Learning Representations

    2018

  2. [10]

    Jiatao Gu and Xiang Kong. 2021. https://aclanthology.org/2021.findings-acl.11 Fully non-autoregressive neural machine translation: Tricks of the trade . In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 120--133

    2021

  3. [11]

    Chenyang Huang, Fei Huang, Zaixiang Zheng, Osmar Za \" ane, Hao Zhou, and Lili Mou. 2023 a . https://aclanthology.org/2023.findings-ijcnlp.14 Multilingual non-autoregressive machine translation without knowledge distillation . In Findings of the Association for Computational L...

    2023

  4. [12]

    Chenyang Huang, Wei Yang, Yanshuai Cao, Osmar Za \" ane, and Lili Mou. 2021. https://aclanthology.org/2021.spnlp-1.7 A globally normalized neural model for semantic parsing . In Proceedings of the 5th Workshop on Structured Prediction for NLP, pages 61--66

    2021

  5. [13]

    Chenyang Huang, Hao Zhou, Osmar R Za \" ane, Lili Mou, and Lei Li. 2022 a . https://ojs.aaai.org/index.php/AAAI/article/view/21323 Non-autoregressive translation with layer-wise prediction and deep supervision . In Proceedings of the AAAI Conference on Artificial Intelligence,...

    2022

  6. [14]

    Fei Huang, Pei Ke, and Minlie Huang. 2023 b . https://doi.org/10.1162/tacl\_a\_00582 Directed acyclic transformer pre-training for high-quality non-autoregressive text generation . Transactions of the Association for Computational Linguistics, 11:941--959

    2023 doi

  7. [15]

    Fei Huang, Hao Zhou, Yang Liu, Hang Li, and Minlie Huang. 2022 b . https://icml.cc/virtual/2022/spotlight/17958 Directed acyclic transformer for non-autoregressive machine translation . In International Conference on Machine Learning, pages 9410--9428

    2022

  8. [16]

    Koller and N

    D. Koller and N. Friedman. 2009. https://books.google.co.in/books?id=7dzpHCHzNQ4C Probabilistic Graphical Models: Principles and Techniques . MIT Press

    2009

  9. [17]

    Lafferty, Andrew McCallum, and Fernando C

    John D. Lafferty, Andrew McCallum, and Fernando C. N. Pereira. 2001. https://dl.acm.org/doi/10.5555/645530.655813 Conditional random fields: P robabilistic models for segmenting and labeling sequence data . In Proceedings of the Eighteenth International Conference on Machine L...

    2001

  10. [18]

    Ann Lee, Michael Auli, and Marc ' Aurelio Ranzato. 2021. https://aclanthology.org/2021.acl-long.563 Discriminative reranking for neural machine translation . In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International J...

    2021

  11. [19]

    Jason Lee, Elman Mansimov, and Kyunghyun Cho. 2018. https://www.aclweb.org/anthology/D18-1149 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

  12. [20]

    Chin-Yew Lin. 2004. https://aclanthology.org/W04-1013 ROUGE : A package for automatic evaluation of summaries . In Text Summarization Branches Out, pages 74--81

    2004

  13. [21]

    Puyuan Liu, Chenyang Huang, and Lili Mou. 2022 a . https://aclanthology.org/2022.acl-long.545 Learning non-autoregressive models from search for unsupervised sentence summarization . In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Vo...

    2022

  14. [22]

    Puyuan Liu, Xiang Zhang, and Lili Mou. 2022 b . https://openreview.net/forum?id=KXybrIUJnya A character-level length-control algorithm for non-autoregressive sentence summarization . In Advances in Neural Information Processing Systems, pages 29101--29112

    2022

  15. [23]

    Yizhu Liu, Zhiyi Luo, and Kenny Zhu. 2018. https://aclanthology.org/D18-1444 Controlling length in abstractive summarization using a convolutional neural network . In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 4110--4119

    2018

  16. [24]

    Clara Meister, Ryan Cotterell, and Tim Vieira. 2020. https://aclanthology.org/2020.emnlp-main.170 If beam search is the answer, what was the question? In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 2173--2185

    2020

  17. [25]

    Ani Nenkova, Sameer Maskey, and Yang Liu. 2011. https://aclanthology.org/volumes/P11-5/ Automatic summarization . In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Tutorial Abstracts, page 3

    2011

  18. [26]

    Franz Josef Och, Daniel Gildea, Sanjeev Khudanpur, Anoop Sarkar, Kenji Yamada, Alex Fraser, Shankar Kumar, Libin Shen, David Smith, Katherine Eng, Viren Jain, Zhen Jin, and Dragomir Radev. 2004. https://aclanthology.org/N04-1021 A smorgasbord of features for statistical machin...

    2004

  19. [27]

    Lihua Qian, Hao Zhou, Yu Bao, Mingxuan Wang, Lin Qiu, Weinan Zhang, Yong Yu, and Lei Li. 2021. https://aclanthology.org/2021.acl-long.155 Glancing transformer for non-autoregressive neural machine translation . In Proceedings of the 59th Annual Meeting of the Association for C...

    2021

  20. [28]

    Mathieu Ravaut, Shafiq Joty, and Nancy Chen. 2022. https://aclanthology.org/2022.acl-long.309 S umma R eranker: A multi-task mixture-of-experts re-ranking framework for abstractive summarization . In Proceedings of the 60th Annual Meeting of the Association for Computational L...

    2022

  21. [29]

    Rush, Sumit Chopra, and Jason Weston

    Alexander M. Rush, Sumit Chopra, and Jason Weston. 2015. https://aclanthology.org/D15-1044 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]

    Raphael Schumann, Lili Mou, Yao Lu, Olga Vechtomova, and Katja Markert. 2020. https://aclanthology.org/2020.acl-main.452 Discrete optimization for unsupervised sentence summarization with word-level extraction . In Proceedings of the 58th Annual Meeting of the Association for ...

    2020

  23. [31]

    Chenze Shao and Yang Feng. 2022. https://openreview.net/forum?id=Qvh0SAPrYzH Non-monotonic latent alignments for CTC -based non-autoregressive machine translation . In Advances in Neural Information Processing Systems, pages 8159--8173

    2022

  24. [32]

    Chenze Shao, Zhengrui Ma, and Yang Feng. 2022. https://aclanthology.org/2022.findings-emnlp.322 V iterbi decoding of directed acyclic transformer for non-autoregressive machine translation . In Findings of the Association for Computational Linguistics: EMNLP 2022, pages 4390--4397

    2022

  25. [33]

    Chenhui Shen, Liying Cheng, Xuan-Phi Nguyen, Yang You, and Lidong Bing. 2023. https://aclanthology.org/2023.findings-emnlp.278 Large language models are not yet human-level evaluators for abstractive summarization . In Findings of the Association for Computational Linguistics:...

    2023

  26. [34]

    Felix Stahlberg and Bill Byrne. 2019. https://aclanthology.org/D19-1331 On NMT search errors and model errors: Cat got your tongue? In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural L...

    2019

  27. [35]

    Yixuan Su, Deng Cai, Yan Wang, David Vandyke, Simon Baker, Piji Li, and Nigel Collier. 2021. https://aclanthology.org/2021.eacl-main.18 Non-autoregressive text generation with pre-trained language models . In Proceedings of the 16th Conference of the European Chapter of the As...

    2021

  28. [36]

    Ilya Sutskever, Oriol Vinyals, and Quoc V. Le. 2014. https://proceedings.neurips.cc/paper/2014/hash/a14ac55a4f27472c5d894ec1c3c743d2-Abstract.html Sequence to sequence learning with neural networks . In Advances in Neural Information Processing Systems, pages 3104--3112

    2014

  29. [37]

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. 2016. https://doi.org/10.1109/CVPR.2016.308 Rethinking the inception architecture for computer vision . In 2016 IEEE Conference on Computer Vision and Pattern Recognition, pages 2818--2826

    2016 doi

  30. [38]

    Sho Takase and Naoaki Okazaki. 2019. https://aclanthology.org/N19-1401 Positional encoding to control output sequence length . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Vo...

    2019

  31. [39]

    Akim Tsvigun, Ivan Lysenko, Danila Sedashov, Ivan Lazichny, Eldar Damirov, Vladimir Karlov, Artemy Belousov, Leonid Sanochkin, Maxim Panov, Alexander Panchenko, Mikhail Burtsev, and Artem Shelmanov. 2022. https://aclanthology.org/2022.findings-emnlp.377 Active learning for abs...

    2022

  32. [40]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. 2017. https://proceedings.neurips.cc/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf Attention is all you need . In Advances in Neural Informa...

    2017

  33. [41]

    Yuqiao Wen, Behzad Shayegh, Chenyang Huang, Yanshuai Cao, and Lili Mou. 2024. https://arxiv.org/abs/2403.00144 EBBS : An ensemble with bi-level beam search for zero-shot machine translation . arXiv preprint arXiv:2403.00144

    2024 arXiv

  34. [42]

    Wen Zhang, Yang Feng, Fandong Meng, Di You, and Qun Liu. 2019. https://aclanthology.org/P19-1426 Bridging the gap between training and inference for neural machine translation . In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4...

    2019

  35. [43]

    Chujie Zheng, Hao Zhou, Fandong Meng, Jie Zhou, and Minlie Huang. 2023. https://openreview.net/forum?id=shr9PXz7T0 Large language models are not robust multiple choice selectors . In International Conference on Learning Representations

    2023

  36. [44]

    Liu Zhuang, Lin Wayne, Shi Ya, and Zhao Jun. 2021. https://aclanthology.org/2021.ccl-1.108 A robustly optimized BERT pre-training approach with post-training . In Proceedings of the 20th Chinese National Conference on Computational Linguistics, pages 1218--1227

    2021

  37. [45]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  38. [46]

    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 gl...

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.