Pith. sign in

REVIEW 3 major objections 5 minor 34 references

Improving Low-Resource Translation with Dictionary-Guided Fine-Tuning and RL: A Spanish-to-Wayuunaiki Study

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A small LLM that can look words up in a bilingual dictionary and learns when to do so via RL translates Spanish-Wayuunaiki better than supervised fine-tuning alone.

desk verdict A genuinely new RL-plus-dictionary recipe for low-resource MT, with thoughtful ablations but effect sizes that sit on single runs; deserves review, not a desk reject. read the letter →

arxiv 2508.19481 v1 pith:DQIO6QPL submitted 2025-08-26 cs.CL cs.AI

Manuel Mosquera , Melissa Robles , Johan Rodriguez , Ruben Manrique This is my paper · ORCID
classification cs.CLcs.AI
keywords low-resourcemachinetranslationWayuunaikidictionary-guidedGRPOreinforcementlearningtool-augmentedLLMSpanish-WayuunaikiBLEU
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

Wayuunaiki, an Arawakan language spoken by about 420,000 people in Colombia and Venezuela, has very little parallel text for machine translation, so the paper asks whether an external bilingual dictionary can make up for missing data. The authors' answer is to treat translation as a tool-augmented decision problem: an instruction-tuned language model is first fine-tuned on parallel sentences and synthetic dictionary-call examples, then trained with GRPO reinforcement learning using sentence-level BLEU as the reward. On the AmericasNLP 2025 Spanish-Wayuunaiki test set, their best 0.5B-parameter model reaches 3.42 BLEU, an 11% gain over supervised fine-tuning alone and an 18% relative gain over a supervised baseline without dictionary access, while a 7B variant reaches 4.45 BLEU. The paper presents this as the first application of RL-based tool use to low-resource machine translation, and argues the recipe generalizes to any language pair with a dictionary.

What carries the argument

The load-bearing mechanism is the dictionary tool-augmented generation loop. During decoding the model can emit structured calls of the form <spa to wayuu>word</spa to wayuu>, the environment returns the dictionary's first five matches inside <matches> tags, and the final translation must appear inside <answer> tags. SFT first teaches the model this format using synthetic demonstrations where 0–4 Spanish words are randomly queried, and then GRPO samples 8 candidate translations per prompt, masks tool outputs from the policy loss, and rewards final answers with sentence-level BLEU. This two-stage design lets RL spend its budget on refining when and how to query rather than learning the tool from scratch.

What would settle it

Re-running the Qwen-0.5B SFT and SFT+RL configurations several times with different random seeds and checking whether the dictionary-enabled RL model beats SFT alone in every run would settle whether the 3.42 versus 3.08 BLEU difference is systematic.

Watch

Extended reading notes

Core claim

The central discovery is that the dictionary tool and the RL stage combine to produce a model that uses the lexicon aggressively and effectively. The Qwen2.5-0.5B-Instruct model trained with SFT plus 1400 GRPO steps calls the dictionary in every answer, averaging 3.94 of 4 allowed lookups, with a 95% success rate, and improves BLEU from 3.08 (SFT only) to 3.42. The same pipeline lifts LLaMA-3.2-1B from 3.15 to 3.48 and Qwen2.5-7B to 4.45 BLEU, while the non-instruction-tuned NLLB model, which cannot use the tool, stays at 0.93 BLEU through RL. The paper also reports that the dictionary contributes roughly 6% relative BLEU at both stages, that BLEU as reward works while a character-level edit reward degrades performance by 10.4%, and that the trained model's output is statistically significantly better than picking the dictionary's best suggestion, showing the model combines lexical matches with learned grammar rather than copying.

Load-bearing premise

The paper's comparisons rest on a single successful training run per configuration, so the central assumption is that the reported BLEU gaps—especially the 11% RL improvement—are bigger than run-to-run variation.

Editorial extensions

If this is right

  • The recipe transfers across instruction-tuned architectures: applying the full pipeline to LLaMA-3.2-1B and Qwen2.5-7B improves their BLEU scores, whereas RL alone does not help the non-tool NLLB model.
  • Expected gains on other language pairs are method-dependent: for agglutinative languages like Wayuunaiki the dictionary helps modestly, and the authors argue non-agglutinative languages could benefit more because words translate more independently.
  • Training duration and reward signal matter: the 0.5B model improves from 3.16 after 400 RL steps to 3.42 after 1400 steps, while the CharacTer reward degrades quality even at 1400 steps.
  • A dictionary usable during training does not require high coverage: only 10.4% of unique Spanish test words appear as entries, and the model still gains, suggesting RL can squeeze value from sparse lexical resources.
  • The best configurations use the dictionary on nearly every example, which in deployment would mean predictable tool-call load: up to four lookups per sentence, with most returning valid matches.

Reading between the lines

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

  • A natural next experiment this paper does not run is to vary dictionary coverage deliberately, such as subsampling the 29,000 entries and measuring BLEU gain per lookup, which would separate the value of the tool from the value of RL in a way the current ablations do not.
  • The dictionary's low match rate (only 16.3% of entries that cover a test word supply the reference translation) suggests the method may work even better with a morphological analyzer for Wayuunaiki's agglutinative forms; that is my extrapolation, not a claim in the paper.
  • Because the test set is a single book excerpt with one reference, the 18% relative gain could shrink or grow on more varied contemporary text; a community-sourced evaluation with multiple references and native-speaker judgment would be the test of whether the BLEU gains are usable translations.
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 proposes a tool-augmented translation pipeline for Spanish-to-Wayuunaiki, combining supervised fine-tuning (SFT) on parallel data with synthetic dictionary-use demonstrations, followed by GRPO reinforcement learning (RL) in which sentence-level BLEU against references serves as the reward. The base model is Qwen2.5-0.5B-Instruct, with additional experiments on Llama-3.2-1B, Qwen2.5-7B, and an NLLB checkpoint. The authors report that SFT+RL with dictionary access improves average BLEU from 3.08 (SFT) to 3.42 for Qwen-0.5B, that dictionary access adds roughly 6% relative BLEU, that Qwen-7B+SFT+RL reaches 4.45 BLEU, and that BLEU as a reward outperforms a character-level reward. The evaluation uses the 503-sentence AmericasNLP 2025 shared task test set. The paper also analyzes dictionary lookup success rates and shows that trained models outperform dictionary-only retrieval by a statistically significant margin.

Significance. If the reported gains are robust, the paper makes a useful empirical contribution: it is one of the first demonstrations of RL-based tool use for low-resource machine translation, it uses an official shared-task test set rather than a self-constructed one, and it includes ablations across model architectures, model sizes, reward types, and RL durations. The release of code and data is also valuable for reproducibility. However, the central quantitative claims currently rest on single-run comparisons with small BLEU differences, and BLEU is used both as the RL reward and as the evaluation metric. These issues do not invalidate the approach, but they make the headline improvements provisional until the measurement uncertainty is quantified.

major comments (3)
  1. [Appendix A3, Table 1, Figure 2] The central claim that SFT+RL and dictionary access improve translation quality rests on very small BLEU differences measured from single runs. Appendix A3 states that 'Only one successful run was considered for each experiment,' and Tables 1-3 report no error bars or repeated seeds. The headline comparison, SFT 3.08 vs. SFT+RL 3.42 BLEU, is a 0.34-point (11% relative) difference, and the tool-vs-no-tool comparison is about 6% relative. On a 503-sentence test set, sentence-level BLEU has high variance, and LoRA initialization, RL sampling, or decoding settings could plausibly shift scores by this amount. The paired t-tests in Section 4.1 compare model outputs against dictionary-only retrieval, not SFT vs. SFT+RL, so they do not address whether the RL gain is significant. I would need repeated runs with different seeds, bootstrap confidence intervals over test sentences, or paired significance tests for the specific configuration comparisons that support the abstract's '18% relative gain' and RL-improvement claims.
  2. [Section 3.2, Section 4] BLEU is used both as the RL reward signal and as the evaluation metric. Because the policy is explicitly optimized against sentence-level BLEU, the reported test-set BLEU gains partly reflect better optimization of the training objective rather than independent evidence of improved translation quality. The paper's conclusion that the method 'improves translation quality' would be strengthened by reporting complementary automatic metrics (e.g., chrF, COMET) or at least a qualitative analysis of outputs. The authors acknowledge the lack of native-speaker evaluation in the Limitations section, but complementary metrics are feasible even without a native speaker and would help establish that the gains are not an artifact of reward-evaluation coupling.
  3. [Abstract, Section 4] The abstract's claim of 'up to +3.37 BLEU improvement over previous work' is not traceable to a specific comparison in the results. Section 4 states that the prior NLLB-based system achieved 0.93 BLEU on the curated test set, while the best model here (Qwen-7B+SFT+RL) reaches 4.45 BLEU, which would be a +3.52 difference, not +3.37. If the +3.37 refers to a different baseline or a different model configuration, that comparison should be explicitly identified; if it is derived from Table 2 or Figure 2, the exact numbers and the name of the prior system should be stated. As written, the headline quantitative claim is unverifiable from the tables.
minor comments (5)
  1. [Abstract, Section 2] The expansion of GRPO is inconsistent: the abstract says 'Guided Reward Policy Optimization,' while Section 2 calls it 'Generalized Reinforcement Policy Optimization,' and the standard name is 'Group Relative Policy Optimization.' Please standardize the terminology.
  2. [Table 3, References] The CharacTer reward is cited to Morris et al. (2004) on WER/MER, but CharacTer is a character-level translation edit rate metric introduced in a different paper (Wang et al., 2016, 'CharacTer: Translation Edit Rate on Character Level'). The current reference does not support the metric used.
  3. [Section 4.1] The text says the SFT model's mean BLEU 'reaches 3.07,' but Table 1 reports 3.08 for the same configuration. Please align the numbers.
  4. [Section 2] The statements that 'RL methods have yet to be explored in the context of machine translation' and that this is 'the first to apply RL to adapt LLMs in the translation context' are contradicted by the paper's own citation of Zhan et al. (2024) and by broader prior work on RL for MT. Please soften these overclaims.
  5. [Figure 2] The caption says the dictionary tool provides an 'estimated 6% relative gain,' but Figure 2 does not show error bars or confidence intervals. Given the single-run setup, the caption should describe this as an observed difference in one run rather than an estimate with implied precision.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported gains are measured against an external held-out test set, and BLEU's dual role as reward and metric is not a derivation shortcut.

full rationale

The paper's central claim is an empirical comparison on the AmericasNLP 2025 Shared Task test set, a 503-sentence held-out dataset of native Wayuunaiki translations, which is external to the training corpus and not a fitted parameter. No equation in the paper defines the reported BLEU gains in terms of the training inputs; the SFT and GRPO procedures are standard training methods, and the reported improvements are measured on the test set rather than derived from the training objective. The main self-referential elements are citations to the authors' own prior work for the parallel corpus [22], the prior translation system baseline [25], and the shared-task findings [3]; these provide datasets and comparison numbers, but they are not invoked as uniqueness theorems, ansatz justifications, or load-bearing arguments. BLEU appears both as the GRPO reward and as the evaluation metric, but evaluation is on a held-out test set, so the observed gains are not guaranteed by construction; this is metric alignment common in MT, not circular reasoning. The single-run limitation in Appendix A3 is a statistical robustness concern, not a circularity one, and the absence of significance tests does not make the claim definitionally equivalent to its inputs.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The method does not introduce a mathematical derivation or fitted constants. It rests on empirical assumptions: that the base models can learn Wayuunaiki from limited data, that BLEU against a single reference is a meaningful reward and metric for this agglutinative language, that the random SFT tool demonstrations transfer to useful RL behavior, and that the external dictionary is accurate enough to help. The hand-chosen quantities that directly shape the results are the cap of four tool calls per sample, the five-word filter on dictionary entries, and the random 0-4 demonstration count.

free parameters (3)
  • max_tool_calls_per_sample = 4
    Manually capped at 4 for cost efficiency (Section 3.5); constrains tool usage and sets the theoretical upper bound of successful queries at 1,798.
  • dictionary_filter_max_spanish_words = 5
    Entries with more than five Spanish words were removed, shrinking the dictionary from about 74,000 to about 29,000 entries (Section 3.3), which directly affects tool recall and output quality.
  • synthetic_demonstration_count = 0 to 4 per sample
    Randomly chosen number of dictionary-lookup demonstrations inserted into SFT prompts (Section 3.1); chosen by hand, not optimized, but it shapes the tool-use behavior learned before RL.
assumptions (5)
  • domain assumption The AmericasNLP 2025 test set with single-reference BLEU is a valid measure of translation quality for Wayuunaiki.
    Used for both RL reward and evaluation; the authors acknowledge BLEU does not capture fluency or cultural appropriateness (Section 7).
  • domain assumption Qwen2.5-0.5B-Instruct and the other tested models can acquire enough Wayuunaiki from 59,715 paired sentences plus dictionary access to produce meaningful translations.
    The entire method depends on the base model being capable of learning an agglutinative Indigenous language from small data.
  • ad hoc to paper Randomly generated dictionary-use demonstrations in SFT teach useful tool-use habits that RL can refine.
    Section 3.1 states these examples are 'probably useless' for the correct translation but are relied on to teach the tool-calling format before RL.
  • standard math GRPO with a relaxed KL penalty (following DAPO and Dr.GRPO) is a stable optimizer for this setting.
    Adopted from prior literature (Section 3.2); no convergence guarantees are provided for this task.
  • domain assumption The bilingual dictionary by Negrette Amaya is a reliable lexical resource for Wayuunaiki-Spanish.
    The tool's output is only as good as the dictionary; the paper reports only 10.4% coverage of test words and 16.3% match rate (Section 4.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Low-Resource Translation with Dictionary-Guided Fine-Tuning and RL: A Spanish-to-Wayuunaiki Study." pith.science (2026). https://pith.science/paper/DQIO6QPL

@misc{pith2026250819481,
  author       = {Pith},
  title        = {Pith review of: Improving Low-Resource Translation with Dictionary-Guided Fine-Tuning and RL: A Spanish-to-Wayuunaiki Study},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DQIO6QPL}},
  note         = {Machine review of arXiv:2508.19481}
}
read the original abstract

Low-resource machine translation remains a significant challenge for large language models (LLMs), which often lack exposure to these languages during pretraining and have limited parallel data for fine-tuning. We propose a novel approach that enhances translation for low-resource languages by integrating an external dictionary tool and training models end-to-end using reinforcement learning, in addition to supervised fine-tuning. Focusing on the Spanish-Wayuunaiki language pair, we frame translation as a tool-augmented decision-making problem in which the model can selectively consult a bilingual dictionary during generation. Our method combines supervised instruction tuning with Guided Reward Policy Optimization (GRPO), enabling the model to learn both when and how to use the tool effectively. BLEU similarity scores are used as rewards to guide this learning process. Preliminary results show that our tool-augmented models achieve up to +3.37 BLEU improvement over previous work, and a 18% relative gain compared to a supervised baseline without dictionary access, on the Spanish-Wayuunaiki test set from the AmericasNLP 2025 Shared Task. We also conduct ablation studies to assess the effects of model architecture and training strategy, comparing Qwen2.5-0.5B-Instruct with other models such as LLaMA and a prior NLLB-based system. These findings highlight the promise of combining LLMs with external tools and the role of reinforcement learning in improving translation quality in low-resource language settings.

Figures

Figures reproduced from arXiv: 2508.19481 by the authors.

Figure 1
Figure 1. Overview of the training pipeline. A large language model is first finetuned using supervised learning on Span [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Average BLEU scores for different Qwen model variants, with and without tool usage. The results show that [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Number of dictionary lookups that returned re [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 20 canonical work pages

  1. [1]

    Osf spanish-wayuunaki,

    Rafael Jose Negrette Amaya. Osf spanish-wayuunaki,

    work page

  2. [2]

    Are ai agents the new machine translation frontier? challenges and opportunities of single- and multi-agent systems for multilingual digital communication, 2025

    Vicent Briva-Iglesias. Are ai agents the new machine translation frontier? challenges and opportunities of single- and multi-agent systems for multilingual digital communication, 2025. URL https://arxiv.org/abs/ 2504.12891

    arXiv 2025

  3. [3]

    Ortega, Luis Chiruzzo, Arturo On- cevay, Shruti Rijhwani, Katharina Von Der Wense, and Manuel Mager

    Ona De Gibert, Robert Pugh, Ali Marashian, Raul Vazquez, Abteen Ebrahimi, Pavel Denisov, Enora Rice, Edward Gow-Smith, Juan Prieto, Melissa Rob- les, Rub´ en Manrique, Oscar Moreno, Angel Lino, Rolando Coto-Solano, Aldo Alvarez, Marvin Ag¨ uero- Torales, John E. Ortega, Luis Chiruzzo, Arturo On- cevay, Shruti Rijhwani, Katharina Von Der Wense, and Manuel ...

    work page 2025

  4. [4]

    DeepSeek-AI, Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhi- hong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, Bing Xue, Bingxuan Wang, Bochao Wu, Bei Feng, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, Damai...

    arXiv 2025

  5. [5]

    Retool: Reinforcement learn- ing for strategic tool use in llms, 2025

    Jiazhan Feng, Shijue Huang, Xingwei Qu, Ge Zhang, Yujia Qin, Baoquan Zhong, Chengquan Jiang, Jinxin Chi, and Wanjun Zhong. Retool: Reinforcement learn- ing for strategic tool use in llms, 2025. URL https: //arxiv.org/abs/2504.11536

    arXiv 2025

  6. [6]

    Kanishk Gandhi, Ayush Chakravarthy, Anikait Singh, Nathan Lile, and Noah D. Goodman. Cognitive be- haviors that enable self-improving reasoners, or, four habits of highly effective stars, 2025. URL https: //arxiv.org/abs/2503.01307

    arXiv 2025

  7. [7]

    Anna Goldie, Azalia Mirhoseini, Hao Zhou, Irene Cai, and Christopher D. Manning. Synthetic data genera- tion and multi-step rl for reasoning and tool use, 2025. URL https://arxiv.org/abs/2504.04736

    arXiv 2025

  8. [8]

    Enriching Way´ uunaiki-Spanish neu- ral machine translation with linguistic information

    Nora Graichen, Josef Van Genabith, and Cristina Espa˜ na-bonet. Enriching Way´ uunaiki-Spanish neu- ral machine translation with linguistic information. In Manuel Mager, Abteen Ebrahimi, Arturo Once- vay, Enora Rice, Shruti Rijhwani, Alexis Palmer, and Katharina Kann, editors, Proceedings of the Work- shop on Natural Language Processing for Indigenous Lan...

    doi:10.18653/v1/202 2023

Show all 34 references
  1. [9]

    Proceedings of the First Workshop on Language Models for Low-Resource Languages, Abu Dhabi, United Arab Emirates, January 2025

    Hansi Hettiarachchi, Tharindu Ranasinghe, Paul Rayson, Ruslan Mitkov, Mohamed Gaber, Damith Premasiri, Fiona Anting Tan, and Lasitha Uyan- godage, editors. Proceedings of the First Workshop on Language Models for Low-Resource Languages, Abu Dhabi, United Arab Emirates, January...

    2025

  2. [10]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large lan- guage models, 2021. URL https://arxiv.org/abs/ 2106.09685

    2021 arXiv

  3. [11]

    Machine translation using grammar materials for LLM post-correction

    Jonathan Hus, Antonios Anastasopoulos, and Nathaniel Krasner. Machine translation using grammar materials for LLM post-correction. In Manuel Mager, Abteen Ebrahimi, Robert Pugh, Shruti Rijhwani, Katharina Von Der Wense, Luis Chiruzzo, Rolando Coto-Solano, and Arturo On- cevay,...

    2025

  4. [12]

    Nwatu, Veron- ica Perez-Rosas, Siqi Shen, Zekun Wang, Winston Wu, and Rada Mihalcea

    Oana Ignat, Zhijing Jin, Artem Abzaliev, Laura Bi- ester, Santiago Castro, Naihao Deng, Xinyi Gao, Aylin Ece Gunal, Jacky He, Ashkan Kazemi, Muham- mad Khalifa, Namho Koh, Andrew Lee, Siyang Liu, Do June Min, Shinka Mori, Joan C. Nwatu, Veron- ica Perez-Rosas, Siqi Shen, Zekun...

    2024

  5. [13]

    Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning, March 2025

    Bowen Jin, Hansi Zeng, Zhenrui Yue, Dong Wang, Hamed Zamani, and Jiawei Han. Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning, March 2025. URL http: //arxiv.org/abs/2503.09516. arXiv:2503.09516 [cs]

    2025 arXiv

  6. [14]

    Challenges in adapting multilingual LLMs to low-resource lan- guages using LoRA PEFT tuning

    Omkar Khade, Shruti Jagdale, Abhishek Phaltankar, Gauri Takalikar, and Raviraj Joshi. Challenges in adapting multilingual LLMs to low-resource lan- guages using LoRA PEFT tuning. In Kengatharaiyer Sarveswaran, Ashwini Vaidya, Bal Krishna Bal, Sana Shams, and Surendrabikram Tha...

    2025

  7. [15]

    Gon- zalez, Hao Zhang, and Ion Stoica

    Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gon- zalez, Hao Zhang, and Ion Stoica. Efficient mem- ory management for large language model serving with pagedattention. In Proceedings of the ACM SIGOPS 29th Symposium on Operating Sys...

    2023

  8. [16]

    Understanding r1-zero-like training: A critical per- spective, 2025

    Zichen Liu, Changyu Chen, Wenjun Li, Penghui Qi, Tianyu Pang, Chao Du, Wee Sun Lee, and Min Lin. Understanding r1-zero-like training: A critical per- spective, 2025. URL https://arxiv.org/abs/2503. 20783

    2025

  9. [17]

    The performance of artificial intel- ligence in the use of indigenous american languages,

    Miguel Lucas, Alejandro Burgue˜ no, Miguel Carazas, C´ esar Buenadicha S´ anchez, Smeldy Ramirez Rufino, C´ esar Said Rosales Torres, Daniel Korn, Hiwot Tesfaye, and Gretchen Deo. The performance of artificial intel- ligence in the use of indigenous american languages,

  10. [18]

    From wer and ril to mer and wil: improved evaluation mea- sures for connected speech recognition

    Andrew Morris, Viktoria Maier, and Phil Green. From wer and ril to mer and wil: improved evaluation mea- sures for connected speech recognition. 01 2004

    2004

  11. [19]

    No language left behind: Scaling human- centered machine translation, 2022

    NLLBTeam. No language left behind: Scaling human- centered machine translation, 2022

    2022

  12. [20]

    Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Car- roll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, a...

    2022

  13. [21]

    Bleu: A method for automatic evaluation of machine translation

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei- Jing Zhu. Bleu: A method for automatic evaluation of machine translation. In Proceedings of the 40th An- nual Meeting on Association for Computational Lin- guistics, ACL ’02, USA, 2002. Association for Com- putational Linguis...

    2002

  14. [22]

    Translation systems for low-resource colombian indige- nous languages, a first step towards cultural preserva- tion

    Juan Prieto, Cristian Martinez, Melissa Robles, Al- berto Moreno, Sara Palacios, and Rub´ en Manrique. Translation systems for low-resource colombian indige- nous languages, a first step towards cultural preserva- tion. In Manuel Mager, Abteen Ebrahimi, Shruti Rijh- wani, Artu...

    2024

  15. [23]

    Qwen2.5 technical report, 2025

    Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Junyang Lin, Kai Dang, Keming Lu, Keqin Bao, Kexin Yang, ...

    2025 arXiv

  16. [24]

    Deepspeed: System optimizations en- able training deep learning models with over 100 billion parameters

    Jeff Rasley, Samyam Rajbhandari, Olatunji Ruwase, and Yuxiong He. Deepspeed: System optimizations en- able training deep learning models with over 100 billion parameters. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , KDD ...

    2020

  17. [25]

    Mart´ ınez, Juan C

    Melissa Robles, Cristian A. Mart´ ınez, Juan C. Pri- eto, Sara Palacios, and Rub´ en Manrique. Preserving heritage: Developing a translation tool for indigenous dialects. In Proceedings of the 17th ACM Interna- tional Conference on Web Search and Data Mining , WSDM ’24, page 1...

  18. [26]

    Proximal policy optimiza- tion algorithms, 2017

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimiza- tion algorithms, 2017. URL https://arxiv.org/abs/ 1707.06347

    2017 arXiv

  19. [27]

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y. K. Li, Y. Wu, and Daya Guo. Deepseek- math: Pushing the limits of mathematical reasoning in open language models, 2024. URL https://arxiv. org/abs/2402.03300

    2024 arXiv

  20. [28]

    Journey to the Center of the Earth

    Jules Verne. Journey to the Center of the Earth . Voy- ages extraordinaires. Pierre-Jules Hetzel, Paris, 1864. Originally published as Voyage au centre de la Terre

  21. [29]

    Dapo: An open- source llm reinforcement learning system at scale, 2025

    Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Weinan Dai, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Guangming Sheng, Yuxuan Tong, Chi Zhang, Mofan Zhang, Wang Zhang, Hang Zhu, Jinhua Zhu, Jiaze Chen, Jiangjie Chen...

    2025 arXiv

  22. [30]

    A reinforcement learning approach to improve low-resource machine translation leveraging domain monolingual data

    Hongxiao Zhang, Mingtong Liu, Chunyou Li, Yufeng Chen, Jinan Xu, and Ming Zhou. A reinforcement learning approach to improve low-resource machine translation leveraging domain monolingual data. In Nicoletta Calzolari, Min-Yen Kan, Veronique Hoste, Alessandro Lenci, Sakriani Sa...

    2024

  23. [34]

    and is accessible via the Machine Learning for Indigenous Language Preservation project website. 11 Hyperparameter Definition Value max steps Maximum number of examples seen 1400 sims per prompt Simulations to calculate reward per example 8 policy lr Learning rate for the poli...

  24. [2021]

    URL https://osf.io/6kbze/

  25. [2024]

    ISBN 9798400703713

    Association for Computing Machinery. ISBN 9798400703713. doi: 10.1145/3616855.3637828. URL https://doi.org/10.1145/3616855.3637828

  26. [2025]

    URL https://doi.org/10.18235/0013542

    doi

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.