Pith. sign in

REVIEW 4 major objections 5 minor 53 references

In-Domain African Languages Translation Using LLMs and Multi-armed Bandits

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a UCB multi-armed bandit can select the best machine-translation model for in-domain African-language data, outperforming the strongest individual model by an average of 2.68% BLEU, and that the selection works even…

desk verdict A plausible bandit-based model selection idea for low-resource MT, but the central 2.68% claim collapses under scrutiny and the evaluation hygiene is not there. read the letter →

arxiv 2505.15069 v1 pith:VLYDTZL6 submitted 2025-05-21 cs.CL

Pratik Rakesh Singh , Kritarth Prasad , Mohammadi Zaki , Pankaj Wasnik This is my paper · ORCID
classification cs.CL
keywords machinetranslationlow-resourcelanguagesAfricandomainadaptationmulti-armedbanditscontextualmodelselectionBLEU
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a multi-armed bandit, specifically the Upper Confidence Bound algorithm, can pick the best machine-translation model for a given domain and language when in-domain data is scarce. It reports that this selection approach beats the strongest individual model, NLLB, by an average of 2.68% BLEU across three African languages and three domains, and that the same method works when no reference translations exist. The payoff would be a way to adapt to a domain without fine-tuning and without large parallel corpora, which matters for low-resource languages where such data rarely exists.

What carries the argument

The central machinery is the multi-armed bandit framing: the arms are the candidate MT systems (Aya101, NLLB, Madlad, Gemma2, Llama3.1), the context vector for each source sentence is its LaBSE embedding, and the reward is a normalized blend of BLEU and COMET when references are available, or CometKiwi alone in the target-free case. The four algorithms compared, UCB, LinUCB, Neural LinUCB, and Thompson Sampling, balance exploration and exploitation to identify the best arm, and the claim rests on UCB's regret-minimization behavior surviving on 1,000 samples per language-domain pair.

What would settle it

Re-run the procedure with a strictly separated validation set and a fresh 1,000-sentence test set for each language-domain pair, and check whether UCB still beats the best single model by the reported margin with overlapping confidence intervals.

Watch

Extended reading notes

Core claim

The central claim is that model selection for in-domain machine translation can be framed as a contextual multi-armed bandit problem, with each arm being a pretrained NMT system and the context being a LaBSE embedding of the source sentence. The authors show that UCB-driven selection, using a reward that combines BLEU and COMET when gold translations exist or CometKiwi when they do not, identifies the strongest system with only 1,000 exploration sentences and occasionally selects an alternative system that beats the best single model. Averaged over languages and domains, UCB improves on the strongest baseline, NLLB, by 2.68% BLEU, and the paper argues this gives a statistically grounded selection method that avoids fine-tuning and catastrophic forgetting.

Load-bearing premise

The reported advantage assumes the test set is genuinely held out and was not used for hyperparameter tuning or bandit exploration; if the same 1,000 sentences served both roles, the gains are optimistically biased.

Editorial extensions

If this is right

  • If UCB-based selection works as reported, translators can choose the best system for a domain without fine-tuning, preserving general-domain performance and avoiding the need for large in-domain parallel data.
  • The target-free result implies that model selection is feasible for languages where only source text exists, using reference-less quality metrics as the reward signal.
  • The small but consistent gains over the best single model suggest that a pool of diverse pretrained systems can be exploited dynamically, sentence by sentence, rather than committing to one model.
  • The method offers a statistical basis for selection, reducing the risk of choosing a model from a single high-variance BLEU evaluation on limited data.

Reading between the lines

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

  • A direct cross-check would compare UCB selection against the simpler strategy of scoring every model on a small validation set and picking the best once; the paper does not report this baseline, so part of the gain may come from the exploration setup rather than online selection itself.
  • Because the reported test sentences are used during exploration to compute rewards, the BLEU numbers reflect the bandit's decisions on those same sentences; a strictly separated exploration set and a fresh held-out test set would be a stricter evaluation.
  • The bandit framing should transfer to other low-resource language families and to other generation tasks with reference-free quality metrics, such as summarization or dialogue response selection.
  • If the gains survive a stricter held-out evaluation, the approach becomes a practical low-cost alternative to fine-tuning for rapid domain adaptation across many languages.
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.

Referee Report

4 major / 5 minor

Summary. The paper proposes using multi-armed bandit algorithms (UCB, LinUCB, Neural LinUCB, and Thompson Sampling) to select among five pretrained machine-translation systems for English-to-Yoruba, English-to-Swahili, and English-to-Igbo translation in the News, Movies, and Religious domains. Rewards are defined as a weighted combination of BLEU and reference-based COMET when parallel data is available, and as CometKiwi in a target-free scenario. The authors claim that UCB outperforms the best static system, NLLB, by an average of 2.68% BLEU in the parallel-data case and is also the best algorithm in the target-free scenario.

Significance. The problem addressed is relevant: selecting an adequate translation model for low-resource languages and narrow domains without fine-tuning is practically important, and treating model selection as a contextual bandit is a plausible, lightweight strategy. The paper evaluates on three African languages and three domains, includes both target-available and target-free reward settings, and compares several standard bandit algorithms. If the results were established on a properly held-out test set with significance testing, the contribution would be useful for practitioners. As it stands, however, the manuscript does not provide such evidence, and several internal inconsistencies prevent acceptance in its current form.

major comments (4)
  1. [§4 and §5] The evaluation protocol does not establish that the 1,000 sentences scored in Tables 1 and 2 are disjoint from the sentences on which the bandits observed rewards or on which hyperparameters were tuned. Section 4 states, "initial few sentences of validations were used for convergence of algorithm and the algorithm was tested on rest of the remaining sentences," and §5 says performance is evaluated using 1,000 test samples without stating their provenance. If the same sentences were used for exploration rewards (BLEU+COMET in the parallel-data case) and for final BLEU scoring, the comparison against static NLLB is in-sample and the reported 2.68% gain is not a valid estimate. The authors must specify the exact train/validation/test split, state explicitly that no reward was observed on the final test sentences, and report results on a genuinely held-out set.
  2. [Table 2 and §5] The prose claim that UCB performs best in the target-free scenario is contradicted by the table's own averages. Averaging the nine BLEU entries in Table 2 gives Thompson Sampling 20.84, UCB 20.72, LinUCB 20.65, and Neural LinUCB 19.99. In addition, in Table 1, UCB ties NLLB in five of nine cells and is slightly worse in News-Yoruba (9.539 vs. 9.67); the 2.68% average improvement is driven almost entirely by the Religious-Swahili cell (+4.35). The claim of "superior performance across all languages and domains" is therefore an overstatement. The prose should be reconciled with the tables or the claims should be qualified to report a single average gain with appropriate caveats.
  3. [§1, Figure 1(c), and §5] The abstract and Figure 1(c) promise "optimal model selection with high confidence" and statistically significant results, but no significance tests, confidence intervals, or run-to-run variance are reported anywhere in the manuscript. With average differences of a few tenths of a BLEU point over 1,000 sentences, the observed gaps (e.g., News-Igbo +0.10 in Table 1) are within typical noise. The authors should provide paired bootstrap confidence intervals or equivalent significance tests for the headline comparisons, and report the number of independent runs and the variance across runs.
  4. [§3 and §4] The reward design is partly circular with respect to the evaluation metric. When parallel data is present, the reward is λ·BLEU + (1−λ)·COMET, and the same BLEU metric is used for the final evaluation in Tables 1 and 2. Since λ=0.4 was selected by "hit and trial" on the validation set (§4), the comparison is in-sample with respect to the reward design. The authors should either use a reward that excludes the evaluation metric on the test sentences, or demonstrate that the conclusions are stable across a range of λ values and on a strictly held-out test set.
minor comments (5)
  1. [Abstract] The sentence beginning "As a result, selecting an optimal model..." is grammatically incomplete and should be revised; "As a result" should be capitalized.
  2. [Table 3] The hyperparameter table is incomplete: UCB's λ is listed as a range "[0.4-0.6]" while the text says λ=0.4 gives the best results; the exact values and the tuning procedure should be reported for every algorithm.
  3. [Tables 1 and 2] The tables use inconsistent decimal precision (e.g., 3.847, 7.087, 9.539); all BLEU scores should be reported with the same number of decimal places.
  4. [Figure 1] The figure caption labels panel (c) as "RL algo." and the panels are not clearly referenced in the text; a fuller caption and explicit in-text references would improve readability.
  5. [References] The reference list contains duplicate entries (Saunders 2022a and 2022b) and malformed author names (e.g., "Ricardo et al. Rei", "et al., 2024b"); these should be cleaned up.

Circularity Check

1 steps flagged · score 7.0 of 10

The reported UCB gain is partially in-sample: the bandit's reward (BLEU+COMET) is explored on the same validation-derived sentences later scored as 'test' BLEU, so the selection is fitted to the evaluation metric on the evaluation set.

  1. fitted input called prediction [Section 4 'Choice of Hyper-parameters' and Section 5 'When parallel data is present']
    "We sample 1,000 parallel samples for validation (seed data for model convergence) and testing each. ... initial few sentences of validations were used for convergence of algorithm and the algorithm was tested on rest of the remaining sentences. ... The Bandit-based models explored the reward that is a weighted summation of BLEU and reference-based Comet on the exploration set (Section 3 Rewards), the performance is evaluated using 1000 test samples per domain and language."

    The same 1,000-sample validation set is the source of both the exploration data and the reported test data. The bandit's reward is Reward=λ·BLEU+(1−λ)·COMET, and the paper evaluates the selection with BLEU on 'the rest of the remaining sentences.' If the bandit updates on rewards from those sentences (as 'on-the-fly' selection implies), then the arm choices are fitted to the BLEU scores that are later reported as out-of-sample predictions. The comparison to static NLLB, which never saw these sentences, is therefore not a held-out evaluation but an in-sample fit. The hyperparameter λ was also tuned on this same validation set, further eroding independence.

full rationale

The central quantitative claim—UCB outperforms NLLB by 2.68% average BLEU (§5)—rests on a validation/evaluation protocol that the paper describes ambiguously but in its most literal reading uses the same 1,000 samples for both bandit exploration (reward generation) and final BLEU scoring. Since the reward includes BLEU, the bandit's selection is partly optimized against the very metric and sentences on which it is judged; this is a fitted-input-called-prediction circularity. The static baselines receive no such in-sample adaptation, so the reported improvement is not a clean out-of-sample result. The bandit algorithms themselves (UCB, LinUCB, TS, NL) are external and well-established, so the methodology framework is not circular. The self-citation to Prasad et al. (2025) is a related-work mention and is not load-bearing. A separate internal inconsistency—Table 2's row averages contradict the prose claim that 'UCB performs the best' in the target-free scenario—is a correctness or presentation problem, not a circularity, and is noted here without affecting the circularity score. Given that the main reported gain may be an artifact of in-sample fitting, the circularity score is 7.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

No new entities are introduced. The quantitative claims rest on fitted reward weights, exploration constants, and the assumption that per-sentence reward signals are informative; none of these are derived from first principles, and no external benchmark validates the per-sentence reward.

free parameters (6)
  • reward fusion weight λ = 0.4
    Set by trial on the validation set; §3 Rewards states 'we find that λ=0.4 achieves the best results in our case.' This weight directly defines the reward and therefore the arm rankings.
  • UCB exploration coefficient α = range 0.4-0.6
    Table 3 lists a range with no selection method given. The exploration constant controls how much the algorithm explores and affects model selection.
  • LinUCB exploration coefficient α = 1.5
    Table 3 lists α=1.5 with no tuning procedure; LinUCB also requires a regularization parameter that is not reported.
  • Thompson Sampling prior Beta parameters = Beta(0,0)
    Table 3 sets an improper Haldane prior; this initial belief influences early pulls and is not an output of data.
  • Neural LinUCB network architecture = 2-layer MLP, 50 neurons per layer
    Table 3 gives only the size; no training details, learning rate, or regularization, and the representation affects LinUCB's features.
  • exploration set size = unspecified
    §4 says the algorithm converged on an unspecified number of validation sentences before testing; the exploration budget directly affects which arm is chosen.
assumptions (4)
  • domain assumption Rewards follow a linear function of LaBSE context: r = x^T θ_a + noise (LinUCB)
    Stated in §3 Linear UCB. It is unlikely that per-sentence translation quality is linear in a high-dimensional sentence embedding, but the algorithm's guarantees rely on this.
  • domain assumption LaBSE embeddings contain enough information to distinguish model quality per sentence
    The context vector is the only arm-selection signal; no analysis connects embedding similarities to translation quality differences.
  • domain assumption Sentence-level BLEU and COMET (or CometKiwi) are valid per-sentence rewards
    Rewards are generated per sentence (§3 Rewards); sentence-level BLEU is known to be noisy and often zero for short sentences. The paper normalizes scores into [0,1] but does not validate the per-sentence signal.
  • domain assumption Fixed arm pool contains the best attainable model; selection cannot exceed the pool's ceiling
    The bandit can only choose among five fixed pre-trained systems; 'optimal model selection' is relative to this pool, not to all possible systems.

how reviews work

0 comments
Cite this review

Pith. "Pith review of In-Domain African Languages Translation Using LLMs and Multi-armed Bandits." pith.science (2026). https://pith.science/paper/VLYDTZL6

@misc{pith2026250515069,
  author       = {Pith},
  title        = {Pith review of: In-Domain African Languages Translation Using LLMs and Multi-armed Bandits},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VLYDTZL6}},
  note         = {Machine review of arXiv:2505.15069}
}
read the original abstract

Neural Machine Translation (NMT) systems face significant challenges when working with low-resource languages, particularly in domain adaptation tasks. These difficulties arise due to limited training data and suboptimal model generalization, As a result, selecting an optimal model for translation is crucial for achieving strong performance on in-domain data, particularly in scenarios where fine-tuning is not feasible or practical. In this paper, we investigate strategies for selecting the most suitable NMT model for a given domain using bandit-based algorithms, including Upper Confidence Bound, Linear UCB, Neural Linear Bandit, and Thompson Sampling. Our method effectively addresses the resource constraints by facilitating optimal model selection with high confidence. We evaluate the approach across three African languages and domains, demonstrating its robustness and effectiveness in both scenarios where target data is available and where it is absent.

Figures

Figures reproduced from arXiv: 2505.15069 by the authors.

Figure 1
Figure 1. Motivation for Reinforcement Learning for [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Block diagram of the proposed bandit-based [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 25 canonical work pages

  1. [1]

    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 INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  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]

    Yasin Abbasi-yadkori, D\' a vid P\' a l, and Csaba Szepesv\' a ri. 2011. https://proceedings.neurips.cc/paper_files/paper/2011/file/e1d5be1c7f2f456670de3d53c7b54f4a-Paper.pdf Improved algorithms for linear stochastic bandits . In Advances in Neural Information Processing Systems, volume 24. Curran Associates, Inc

    work page 2011

  4. [4]

    David Adelani et al. 2022. https://doi.org/10.18653/v1/2022.naacl-main.223 A few thousand translations go a long way! leveraging pre-trained models for A frican news translation . In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 3053--3070, Seattle, Uni...

    doi:10.18653/v1/2022.naacl-main.223 2022

  5. [5]

    Peter Auer, Nicol \`o Cesa-Bianchi, and Paul Fischer. 2002. https://doi.org/10.1023/A:1013689704352 Finite-time analysis of the multiarmed bandit problem . Machine Learning, 47(2):235--256

    doi:10.1023/a:1013689704352 2002

  6. [6]

    Seth Aycock and Rachel Bawden. 2024. https://aclanthology.org/2024.eacl-srw.13/ Topic-guided example selection for domain adaptation in LLM -based machine translation . In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: Student Research Workshop, pages 175--195, St. Julian ' s, Malta. Associatio...

    work page 2024

  7. [7]

    Ankur Bapna and Orhan Firat. 2019. https://doi.org/10.18653/v1/D19-1165 Simple, scalable adaptation for neural machine translation . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 1538--1548, Hong Kong, China. Associat...

    doi:10.18653/v1/d19-1165 2019

  8. [8]

    Costa-juss \`a , Fethi Bougares, and Olivier Galibert

    Lo \"i c Barrault, Magdalena Biesialska, Marta R. Costa-juss \`a , Fethi Bougares, and Olivier Galibert. 2020. https://aclanthology.org/2020.wmt-1.2/ Findings of the first shared task on lifelong learning machine translation . In Proceedings of the Fifth Conference on Machine Translation, pages 56--64, Online. Association for Computational Linguistics

    work page 2020

Show all 53 references
  1. [9]

    Djallel Bouneffouf, Irina Rish, and Charu Aggarwal. 2020. https://doi.org/10.1109/CEC48606.2020.9185782 Survey on applications of multi-armed and contextual bandits . In 2020 IEEE Congress on Evolutionary Computation (CEC), pages 1--8

    2020

  2. [10]

    Etienne Boursier and Vianney Perchet. 2024. https://arxiv.org/abs/2211.16275 A survey on multi-player bandits . Preprint, arXiv:2211.16275

    2024 arXiv

  3. [11]

    Guanhua Chen, Yun Chen, Yong Wang, and Victor O. K. Li. 2021. Lexical-constraint-aware neural machine translation via data augmentation. In Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI'20

    2021

  4. [12]

    Mara Chinea-R \'i os, \'A lvaro Peris, and Francisco Casacuberta. 2017. https://doi.org/10.18653/v1/W17-4714 Adapting neural machine translation with parallel synthetic data . In Proceedings of the Second Conference on Machine Translation, pages 138--147, Copenhagen, Denmark. ...

    2017 doi

  5. [13]

    Chenhui Chu and Rui Wang. 2018. https://aclanthology.org/C18-1111/ A survey of domain adaptation for neural machine translation . In Proceedings of the 27th International Conference on Computational Linguistics, pages 1304--1319, Santa Fe, New Mexico, USA. Association for Comp...

    2018

  6. [14]

    Ahmed El-Kishky, Vishrav Chaudhary, Francisco Guzman, and Philipp Koehn. 2020. https://arxiv.org/abs/1911.06154 Ccaligned: A massive collection of cross-lingual web-document pairs . Preprint, arXiv:1911.06154

    2020 arXiv

  7. [15]

    Xavier Garcia, Yamini Bansal, Colin Cherry, George Foster, Maxim Krikun, Fangxiaoyu Feng, Melvin Johnson, and Orhan Firat. 2023. https://arxiv.org/abs/2302.01398 The unreasonable effectiveness of few-shot learning for machine translation . Preprint, arXiv:2302.01398

    2023 arXiv

  8. [16]

    Junjie Hu, Mengzhou Xia, Graham Neubig, and Jaime Carbonell. 2019. https://doi.org/10.18653/v1/P19-1286 Domain adaptation of neural machine translation by lexicon induction . In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 2989...

    2019 doi

  9. [17]

    Dongfu Jiang, Xiang Ren, and Bill Yuchen Lin. 2023. https://doi.org/10.18653/v1/2023.acl-long.792 LLM -blender: Ensembling large language models with pairwise ranking and generative fusion . In Proceedings of the 61st Annual Meeting of the Association for Computational Linguis...

    2023 doi

  10. [18]

    Di Jin, Zhijing Jin, Joey Tianyi Zhou, and Peter Szolovits. 2020. https://arxiv.org/abs/2001.08140 A simple baseline to semi-supervised domain adaptation for machine translation . Preprint, arXiv:2001.08140

    2020 arXiv

  11. [19]

    Catherine Kobus, Josep Crego, and Jean Senellart. 2017. https://arxiv.org/abs/1612.06140 Domain control for neural machine translation . Preprint, arXiv:1612.06140

    2017 arXiv

  12. [20]

    Choquette-Choo, Katherine Lee, Derrick Xin, Aditya Kusupati, Romi Stella, Ankur Bapna, and Orhan Firat

    Sneha Kudugunta, Isaac Caswell, Biao Zhang, Xavier Garcia, Christopher A. Choquette-Choo, Katherine Lee, Derrick Xin, Aditya Kusupati, Romi Stella, Ankur Bapna, and Orhan Firat. 2023. https://arxiv.org/abs/2309.04662 Madlad-400: A multilingual and document-level large audited ...

    2023 arXiv

  13. [21]

    Schapire

    Lihong Li, Wei Chu, John Langford, and Robert E. Schapire. 2010. https://doi.org/10.1145/1772690.1772758 A contextual-bandit approach to personalized news article recommendation . In Proceedings of the 19th International Conference on World Wide Web, WWW '10, page 661–670, New...

    2010

  14. [22]

    Pierre Lison and J \"o rg Tiedemann. 2016. https://aclanthology.org/L16-1147/ O pen S ubtitles2016: Extracting large parallel corpora from movie and TV subtitles . In Proceedings of the Tenth International Conference on Language Resources and Evaluation ( LREC `16) , pages 923...

    2016

  15. [23]

    Yixin Liu and Pengfei Liu. 2021. https://doi.org/10.18653/v1/2021.acl-short.135 S im CLS : A simple framework for contrastive learning of abstractive summarization . In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th Interna...

    2021 doi

  16. [24]

    Zihan Liu, Genta Indra Winata, and Pascale Fung. 2021. https://doi.org/10.18653/v1/2021.findings-acl.239 Continual mixed-language pre-training for extremely low-resource neural machine translation . In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021,...

    2021 doi

  17. [25]

    Tyler Lu, David Pal, and Martin Pal. 2010. https://proceedings.mlr.press/v9/lu10a.html Contextual multi-armed bandits . In Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, volume 9 of Proceedings of Machine Learning Research, pa...

    2010

  18. [26]

    Sreyashi Nag, Mihir Kale, Varun Lakshminarasimhan, and Swapnil Singhavi. 2020. https://arxiv.org/abs/2004.02071 Incorporating bilingual dictionaries for low resource semi-supervised neural machine translation . Preprint, arXiv:2004.02071

    2020 arXiv

  19. [27]

    Khanh Nguyen, Hal Daumé III, and Jordan Boyd-Graber. 2017. https://arxiv.org/abs/1707.07402 Reinforcement learning for bandit neural machine translation with simulated human feedback . Preprint, arXiv:1707.07402

    2017 arXiv

  20. [28]

    Aleksander Obuchowski, Barbara Klaudel, Piotr Frąckowski, Sebastian Krajna, Wasyl Badyra, Michał Czubenko, and Zdzisław Kowalczuk. 2024. https://doi.org/10.1609/aaai.v38i21.30490 Target-free domain adaptation through cross-adaptation (student abstract) . Proceedings of the AAA...

    2024 doi

  21. [29]

    Wong, Shuming Shi, and Zhaopeng Tu

    Jianhui Pang, Fanghua Ye, Longyue Wang, Dian Yu, Derek F. Wong, Shuming Shi, and Zhaopeng Tu. 2024. https://arxiv.org/abs/2401.08350 Salute the classic: Revisiting challenges of machine translation in the age of large language models . Preprint, arXiv:2401.08350

    2024 arXiv

  22. [30]

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. https://doi.org/10.3115/1073083.1073135 Bleu: a method for automatic evaluation of machine translation . In Proceedings of the 40th Annual Meeting on Association for Computational Linguistics, ACL '02, page 311...

    2002

  23. [31]

    Wei Peng, Chongxuan Huang, Tianhao Li, Yun Chen, and Qun Liu. 2020. https://arxiv.org/abs/2004.02577 Dictionary-based data augmentation for cross-domain neural machine translation . Preprint, arXiv:2004.02577

    2020 arXiv

  24. [32]

    MinhQuang Pham, Josep Crego, Fran c ois Yvon, and Jean Senellart. 2019. https://aclanthology.org/2019.iwslt-1.26/ Generic and specialized word embeddings for multi-domain machine translation . In Proceedings of the 16th International Conference on Spoken Language Translation, ...

    2019

  25. [33]

    Alberto Poncelas, Gideon Maillette de Buy Wenniger, and Andy Way. 2019. https://arxiv.org/abs/1906.07808 Adaptation of machine translation models with back-translated data using transductive data selection methods . Preprint, arXiv:1906.07808

    2019 arXiv

  26. [34]

    Matt Post. 2018. https://doi.org/10.18653/v1/W18-6319 A call for clarity in reporting BLEU scores . In Proceedings of the Third Conference on Machine Translation: Research Papers, pages 186--191, Brussels, Belgium. Association for Computational Linguistics

    2018 doi

  27. [35]

    Kritarth Prasad, Mohammadi Zaki, Pratik Singh, and Pankaj Wasnik. 2025. https://arxiv.org/abs/2501.15219 Faster machine translation ensembling with reinforcement learning and competitive correction . Preprint, arXiv:2501.15219

    2025 arXiv

  28. [36]

    Surangika Ranathunga, En-Shiun Annie Lee, Marjana Prifti Skenduli, Ravi Shekhar, Mehreen Alam, and Rishemjit Kaur. 2023. https://doi.org/10.1145/3567592 Neural machine translation for low-resource languages: A survey . ACM Comput. Surv., 55(11)

    2023 doi

  29. [37]

    Mathieu Ravaut, Shafiq Joty, and Nancy F. Chen. 2023. https://arxiv.org/abs/2203.06569 Summareranker: A multi-task mixture-of-experts re-ranking framework for abstractive summarization . Preprint, arXiv:2203.06569

    2023 arXiv

  30. [38]

    Ricardo et al. Rei. 2022. https://aclanthology.org/2022.wmt-1.60/ C omet K iwi: IST -unbabel 2022 submission for the quality estimation shared task . In Proceedings of the Seventh Conference on Machine Translation (WMT), pages 634--645, Abu Dhabi, United Arab Emirates (Hybrid)...

    2022

  31. [39]

    Nguyen, and Katrin Kirchhoff

    Julian Salazar, Davis Liang, Toan Q. Nguyen, and Katrin Kirchhoff. 2020. https://doi.org/10.18653/v1/2020.acl-main.240 Masked language model scoring . In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2699--2712, Online. Associat...

    2020 doi

  32. [41]

    Danielle Saunders. 2022 b . https://arxiv.org/abs/2104.06951 Domain adaptation and multi-domain adaptation for neural machine translation: A survey . Preprint, arXiv:2104.06951

    2022 arXiv

  33. [42]

    Aditya Siddhant, Ankur Bapna, Orhan Firat, Yuan Cao, Mia Xu Chen, Isaac Caswell, and Xavier Garcia. 2022. https://arxiv.org/abs/2201.03110 Towards the next 1000 languages in multilingual machine translation: Exploring the synergy between supervised and self-supervised learning...

    2022 arXiv

  34. [43]

    Pereira, and Leonardo Rocha

    Nícollas Silva, Heitor Werneck, Thiago Silva, Adriano C.M. Pereira, and Leonardo Rocha. 2022. https://doi.org/10.1016/j.eswa.2022.116669 Multi-armed bandits in recommendation systems: A survey of the state-of-the-art and future directions . Expert Systems with Applications, 197:116669

    2022

  35. [44]

    Emmanouil Stergiadis, Satendra Kumar, Fedor Kovalev, and Pavel Levin. 2021. https://aclanthology.org/2021.mtsummit-up.27/ Multi-domain adaptation in neural machine translation through multidimensional tagging . In Proceedings of Machine Translation Summit XVIII: Users and Prov...

    2021

  36. [45]

    NLLB Team. 2022. https://arxiv.org/abs/2207.04672 No language left behind: Scaling human-centered machine translation . Preprint, arXiv:2207.04672

    2022 arXiv

  37. [46]

    Aaron Grattafiori et al. 2024 a . https://arxiv.org/abs/2407.21783 The llama 3 herd of models . Preprint, arXiv:2407.21783

    2024 arXiv

  38. [47]

    Gemma Team et al. 2024 b . https://arxiv.org/abs/2408.00118 Gemma 2: Improving open language models at a practical size . Preprint, arXiv:2408.00118

    2024 arXiv

  39. [48]

    Brian Thompson, Jeremy Gwinnup, Huda Khayrallah, Kevin Duh, and Philipp Koehn. 2019. https://doi.org/10.18653/v1/N19-1209 Overcoming catastrophic forgetting during domain adaptation of neural machine translation . In Proceedings of the 2019 Conference of the North A merican Ch...

    2019 doi

  40. [49]

    William R Thompson. 1933. https://doi.org/10.1093/biomet/25.3-4.285 On the likelihood that one unknown probability exceeds another in view of the evidence of two samples . Biometrika, 25(3-4):285--294

    1933 doi

  41. [50]

    Vaibhav Vaibhav, Sumeet Singh, Craig Stewart, and Graham Neubig. 2019. https://doi.org/10.18653/v1/N19-1190 Improving robustness of machine translation with synthetic noise . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computatio...

    2019 doi

  42. [51]

    Pan Xu, Zheng Wen, Handong Zhao, and Quanquan Gu. 2020. https://arxiv.org/abs/2012.01780 Neural contextual bandits with deep representation and shallow exploration . CoRR, abs/2012.01780

    2020 arXiv

  43. [52]

    Hongxiao Zhang, Hui Huang, Jiale Gao, Yufeng Chen, Jinan Xu, and Jian Liu. 2022. https://aclanthology.org/2022.coling-1.448/ Iterative constrained back-translation for unsupervised domain adaptation of machine translation . In Proceedings of the 29th International Conference o...

    2022

  44. [53]

    Li Zhou. 2016. https://arxiv.org/abs/1508.03326 A survey on contextual multi-armed bandits . Preprint, arXiv:1508.03326

    2016 arXiv

  45. [54]

    Ahmet Üstün, Viraat Aryabumi, Zheng-Xin Yong, Wei-Yin Ko, Daniel D'souza, Gbemileke Onilude, Neel Bhandari, Shivalika Singh, Hui-Lee Ooi, Amr Kayid, Freddie Vargus, Phil Blunsom, Shayne Longpre, Niklas Muennighoff, Marzieh Fadaee, Julia Kreutzer, and Sara Hooker. 2024. https:/...

    2024 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.