---
title: "TranslateGemma for ES-EN Financial Reports: Exploring"
description: "The paper examines whether a small, locally run LLM can be adapted to translate Spanish–English annual reports of IBEX 35… 30 references."
authors: ["Yanco Amor Torterolo Orta", "Melina Chatzi", "Antonio Moreno-Sandoval"]
year: 2026
venue: "The 7th Financial Narrative Processing Workshop"
source: "https://aclanthology.org/2026.fnp-1.8/"
doi: "10.63317/4mfhxunsaksk"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/orta-2026-translategemma-es-en-financial-reports/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/orta-2026-translategemma-es-en-financial-reports/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/46/) / TranslateGemma for ES-EN Financial Reports: Exploring Adaptability to Variable-Sized Contexts

# TranslateGemma for ES-EN Financial Reports: Exploring Adaptability to Variable-Sized Contexts

**Authors:** [Yanco Amor Torterolo Orta](https://md.research.elamur.ai/free/en/authors/yanco-amor-torterolo-orta/), [Melina Chatzi](https://md.research.elamur.ai/free/en/authors/melina-chatzi/) and [Antonio Moreno-Sandoval](https://md.research.elamur.ai/free/en/authors/antonio-moreno-sandoval/) · **Year:** 2026 · **Venue:** [The 7th Financial Narrative Processing Workshop](https://md.research.elamur.ai/free/en/venues/fnp/) ([Financial Narrative Processing Workshop 2026](https://md.research.elamur.ai/free/en/venues/fnp/2026/)) · **Publisher:** [European Language Resources Association (ELRA)](https://md.research.elamur.ai/free/en/publishers/european-language-resources-association-elra/) · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.fnp-1.8/) · [PDF](https://aclanthology.org/2026.fnp-1.8.pdf) · [DOI 10.63317/4mfhxunsaksk](https://doi.org/10.63317/4mfhxunsaksk) · [OpenAlex](https://openalex.org/W7166515497)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper examines whether a small, locally run LLM can be adapted to translate Spanish–English annual reports of IBEX 35 companies in both directions. Local deployment matters to firms because of data privacy and energy cost. The authors fine-tune two TranslateGemma models on a parallel corpus of aligned translation units (TUs). Their data augmentation adds a copy of the corpus in which neighbouring TUs are joined into longer chunks. During training, short and long segments are randomly mixed, and the translation direction is sampled at random. The fine-tuned systems are compared with zero-shot baselines across prompt templates, model sizes and beam search.

The results are split. Fine-tuning raises traditional metric scores for both short and long inputs, but neural metrics prefer the zero-shot outputs. These metrics even rated the human reference below off-the-shelf machine output. The authors read this as a bias of neural metrics toward fluent, literal text, with further weaknesses on long contexts. The low TER of fine-tuned output suggests it fits a company's style closely enough to serve as a draft for human post-editing (MTPE).

## Key points
- **Corpus:** 34 pairs of IBEX 35 annual reports, giving 41,951 ES-EN TUs. These total about 1.26M Spanish and 1.10M English words, and part of the data is released as "Financial ES-EN parallel corpus from annual reports" (Moreno-Sandoval et al., 2025).
- **Dataset versions:**
  - The original version has TUs of at most 697 tokens, split 39,857 / 1,047 / 1,047 (train / val / test).
  - The mixed version adds concatenated chunks of up to 2,102 tokens, bringing the total to 46,076 TUs.
- **Models and training:** google/translategemma-4b-it and google/translategemma-12b-it were tuned with QLoRA (r=16, α=16, 4-bit) through Unsloth. Training ran on a single RTX 5080 with 16GB VRAM, and early stopping ended runs after 2.2–4.1 epochs.
- **Prompts compared:** the built-in Jinja2 template (GP), the leaner Ollama-style prompt (GPO) and a minimal one-line translator instruction.
- **Main finding:** traditional metrics and low TER favour the fine-tuned models, while neural metrics favour zero-shot output. This exposes limitations of current neural metrics for evaluating long-context financial MT.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Universidad Autónoma de Madrid](https://md.research.elamur.ai/free/en/universities/universidad-autonoma-de-madrid/) — University, Madrid, Spain · 7 works · 21 citations
- [Universidad Nacional de Educación a Distancia](https://md.research.elamur.ai/free/en/universities/universidad-nacional-de-educacion-a-distancia/) — University, Madrid, Spain · 6 works · 269 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

## Cite

Cite as: Yanco Amor Torterolo Orta, Melina Chatzi and Antonio Moreno-Sandoval. 2026. TranslateGemma for ES-EN Financial Reports: Exploring Adaptability to Variable-Sized Contexts. The 7th Financial Narrative Processing Workshop.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{orta2026translategemma,
  title = {{TranslateGemma for ES-EN Financial Reports: Exploring Adaptability to Variable-Sized Contexts}},
  author = {Yanco Amor Torterolo Orta, Melina Chatzi and Antonio Moreno-Sandoval},
  booktitle = {The 7th Financial Narrative Processing Workshop},
  year = {2026},
  doi = {10.63317/4mfhxunsaksk},
  url = {https://doi.org/10.63317/4mfhxunsaksk}
}
```

- [← previous Transformers Learning Contrafactives: The Importance of Data Distributions](https://md.research.elamur.ai/free/en/papers/strohmaier-2026-transformers-learning-contrafactives-importance-data/)
- [next → Translation as Augmentation: Effect of Translated Data on Assessment of Difficulty](https://md.research.elamur.ai/free/en/papers/wu-2026-translation-as-augmentation-effect-translated/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/46/)

## References

30 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Emad A. Alghamdi, Jezia Zakraoui, and Fares A. Abanmy. 2023. [Domain adaptation for arabic machine translation: The case of financial texts](https://md.research.elamur.ai/free/en/papers/alghamdi-2023-domain-adaptation-arabic-machine-translation/).
2. Daniel Deutsch, Eleftheria Briakou, Isaac Caswell, Mara Finkelstein, Rebecca Galor, Juraj Juraska, Geza Kovacs, Alison Lui, Ricardo Rei, Jason Riesa, Shruti Rijhwani, Parker Riley, Elizabeth Salesky, Firas Trabelsi, Stephanie Winkler, Biao Zhang, and Markus Freitag. 2025. [Wmt24++: Expanding the lang…](https://md.research.elamur.ai/free/en/papers/deutsch-2025-wmt24-expanding-language-coverage-wmt24/)
3. Christian Herold and Hermann Ney. 2023. On search strategies for document-level neural machine translation. [find](https://scholar.google.com/scholar?q=Christian+Herold+and+Hermann+Ney%2E+2023%2E+On+search+strategies+for+document%2Dlevel+neural+machine+translation%2E)
4. Leticia Herrero Rodes and Verónica Román Mínguez. 2015. English to spanish translation of the economics and finance genres. InTRAlinea: Online Translation Journal, (Special Issue: New Insights into Specialised Translation). Revista del Departamento de Traducción e Interpretación de la Universidad de… [find](https://scholar.google.com/scholar?q=Leticia+Herrero+Rodes+and+Ver%C3%B3nica+Rom%C3%A1n+M%C3%ADnguez%2E+2015%2E+English+to+spanish+translation+of+the+economics+and+finance+genres%2E+InTRAlinea%3A+Online+Translation+Journal%2C+%28Special+Issue%3A+New+Insights+into+Sp)
5. Paolo Di Natale, Elena Chiocchetti, and Egon Waldemar Stemle. 2025. [Meta-evaluation of automatic machine translation metrics between Italian and a minor language variety of German](https://md.research.elamur.ai/free/en/papers/natale-2025-meta-evaluation-automatic-machine-translation/). In Proceedings of the Eleventh Italian Conference on Computational Linguistics (CLiC-it 2025), pages 371–383, Cagliari,…
6. Melvin Johnson, Mike Schuster, Quoc V. Le, Maxim Krikun, Yonghui Wu, Zhifeng Chen, Nikhil Thorat, Fernanda Viégas, Martin Wattenberg, Greg Corrado, Macduff Hughes, and Jeffrey Dean. 2017. [Google’s multilingual neural machine translation system: Enabling zero-shot translation](https://md.research.elamur.ai/free/en/papers/johnson-2017-google-s-multilingual-neural-machine/). Transactions of the Ass…
7. Liang Ding, Di Wu, and Dacheng Tao. 2021. Improving neural machine translation by bidirectional training. [find](https://scholar.google.com/scholar?q=Liang+Ding%2C+Di+Wu%2C+and+Dacheng+Tao%2E+2021%2E+Improving+neural+machine+translation+by+bidirectional+training%2E)
8. Julen Etxaniz, Gorka Azkune, Aitor Soroa, Oier Lopez de Lacalle, and Mikel Artetxe. 2024. [Do multilingual language models think better in English?](https://md.research.elamur.ai/free/en/papers/etxaniz-2024-do-multilingual-language-models-think/) In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume…
9. Juraj Juraska. 2025. Comment on metricx-25 (issue \#12). GitHub Issue Comment. Google Research MetricX Repository. [find](https://scholar.google.com/scholar?q=Juraj+Juraska%2E+2025%2E+Comment+on+metricx%2D25+%28issue+%5C%2312%29%2E+GitHub+Issue+Comment%2E+Google+Research+MetricX+Repository%2E)
10. Juraj Juraska, Daniel Deutsch, Mara Finkelstein, and Markus Freitag. 2024. [MetricX-24: The Google submission to the WMT 2024 metrics shared task](https://md.research.elamur.ai/free/en/papers/juraska-2024-metricx-24-google-submission-wmt/). In Proceedings of the Ninth Conference on Machine Translation, pages 492–504, Miami, Florida, USA. Association for Computational Linguistics.
11. Mara Finkelstein, Isaac Caswell, Tobias Domhan, Jan-Thorsten Peter, Juraj Juraska, Parker Riley, Daniel Deutsch, Geza Kovacs, Cole Dilanni, Colin Cherry, Eleftheria Briakou, Elizabeth Nielsen, Jiaming Luo, Kat Black, Ryan Mullins, Sweta Agrawal, Wenda Xu, Erin Kats, Stephane Jaskiewicz, Markus Freit… [find](https://scholar.google.com/scholar?q=Mara+Finkelstein%2C+Isaac+Caswell%2C+Tobias+Domhan%2C+Jan%2DThorsten+Peter%2C+Juraj+Juraska%2C+Parker+Riley%2C+Daniel+Deutsch%2C+Geza+Kovacs%2C+Cole+Dilanni%2C+Colin+Cherry%2C+Eleftheria+Briakou%2C+Elizabeth+Nielsen%2C+Jiaming)
12. Juraj Juraska, Tobias Domhan, Mara Finkelstein, Tetsuji Nakagawa, Geza Kovacs, Daniel Deutsch, Pidong Wang, and Markus Freitag. 2025. MetricX-25 and GemSpanEval: Google Translate submissions to the WMT25 evaluation shared task. In Proceedings of the Tenth Conference on Machine Translation, pages 957… [find](https://scholar.google.com/scholar?q=Juraj+Juraska%2C+Tobias+Domhan%2C+Mara+Finkelstein%2C+Tetsuji+Nakagawa%2C+Geza+Kovacs%2C+Daniel+Deutsch%2C+Pidong+Wang%2C+and+Markus+Freitag%2E+2025%2E+MetricX%2D25+and+GemSpanEval%3A+Google+Translate+submissions+to+the+WM)
13. Juraj Juraska, Mara Finkelstein, Daniel Deutsch, Aditya Siddhant, Mehdi Mirzazadeh, and Markus Freitag. 2023. [MetricX-23: The Google Submission to the WMT 2023 Metrics Shared Task](https://md.research.elamur.ai/free/en/papers/juraska-2023-metricx-23-google-submission-wmt/). In Proceedings of the Eighth Conference on Machine Translation, pages 756–767, Singapore. Association for Computational…
14. Geza Kovacs, Daniel Deutsch, and Markus Freitag. 2024. Mitigating metric bias in minimum Bayes risk decoding. In Proceedings of the Ninth Conference on Machine Translation, pages 1063–1094, Miami, Florida, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Geza+Kovacs%2C+Daniel+Deutsch%2C+and+Markus+Freitag%2E+2024%2E+Mitigating+metric+bias+in+minimum+Bayes+risk+decoding%2E+In+Proceedings+of+the+Ninth+Conference+on+Machine+Translation%2C+pages+1063%E2%80%931094%2C+Miami%2C+Flo)
15. Zixuan Ke, Yifei Ming, Xuan-Phi Nguyen, Caiming Xiong, and Shafiq Joty. 2025. Demystifying domain-adaptive post-training for financial LLMs. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 31033–31059, Suzhou, China. Association for Computational Ling… [find](https://scholar.google.com/scholar?q=Zixuan+Ke%2C+Yifei+Ming%2C+Xuan%2DPhi+Nguyen%2C+Caiming+Xiong%2C+and+Shafiq+Joty%2E+2025%2E+Demystifying+domain%2Dadaptive+post%2Dtraining+for+financial+LLMs%2E+In+Proceedings+of+the+2025+Conference+on+Empirical+Methods+)
16. Haijun Li, Tianqi Shi, Zifu Shang, Yuxuan Han, Xueyu Zhao, Hao Wang, Yu Qian, Zhiqiang Qian, Linlong Xu, Minghao Wu, Chenyang Lyu, Longyue Wang, Gongbo Tang, Weihua Luo, Zhao Xu, and Kaifu Zhang. 2025. Transbench: Benchmarking machine translation for industrial-scale applications. [find](https://scholar.google.com/scholar?q=Haijun+Li%2C+Tianqi+Shi%2C+Zifu+Shang%2C+Yuxuan+Han%2C+Xueyu+Zhao%2C+Hao+Wang%2C+Yu+Qian%2C+Zhiqiang+Qian%2C+Linlong+Xu%2C+Minghao+Wu%2C+Chenyang+Lyu%2C+Longyue+Wang%2C+Gongbo+Tang%2C+Weihua+Luo%2C+Zhao+Xu%2C+and+Kaifu+Zhang%2E+2025)
17. Tom Kocmi, Ekaterina Artemova, Eleftherios Avramidis, Rachel Bawden, Ondřej Bojar, Konstantin Dranch, Anton Dvorkovich, Sergey Dukanov, Mark Fishel, Markus Freitag, Thamme Gowda, Roman Grundkiewicz, Barry Haddow, Marzena Karpinska, Philipp Koehn, Howard Lakougna, Jessica Lundin, Christof Monz, Kento… — [Findings of the WMT25 General Machine Translation Shared Task: Time to Stop Evaluating on Easy Test Sets](https://md.research.elamur.ai/free/en/papers/kocmi-2025-findings-wmt25-general-machine-translation/)
18. Mariam Nakhlé, Marco Dinarelli, Raheel Qader, Emmanuelle Esperança-Rodier, and Hervé Blanchon. 2025. Dolfin – document-level financial test set for machine translation. [find](https://scholar.google.com/scholar?q=Mariam+Nakhl%C3%A9%2C+Marco+Dinarelli%2C+Raheel+Qader%2C+Emmanuelle+Esperan%C3%A7a%2DRodier%2C+and+Herv%C3%A9+Blanchon%2E+2025%2E+Dolfin+%E2%80%93+document%2Dlevel+financial+test+set+for+machine+translation%2E)
19. Eugene A. Nida. 1964. Toward a Science of Translating. E. J. Brill, Leiden. [find](https://scholar.google.com/scholar?q=Eugene+A%2E+Nida%2E+1964%2E+Toward+a+Science+of+Translating%2E+E%2E+J%2E+Brill%2C+Leiden%2E)
20. Arturo Oncevay, Charese Smiley, and Xiaomo Liu. 2025. The impact of domain-specific terminology on machine translation for finance in European languages. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Tech… [find](https://scholar.google.com/scholar?q=Arturo+Oncevay%2C+Charese+Smiley%2C+and+Xiaomo+Liu%2E+2025%2E+The+impact+of+domain%2Dspecific+terminology+on+machine+translation+for+finance+in+European+languages%2E+In+Proceedings+of+the+2025+Conference+of+the+N)
21. Tom Kocmi, Eleftherios Avramidis, Rachel Bawden, Ondřej Bojar, Anton Dvorkovich, Christian Federmann, Mark Fishel, Markus Freitag, Thamme Gowda, Roman Grundkiewicz, Barry Haddow, Marzena Karpinska, Philipp Koehn, Benjamin Marie, Christof Monz, Kenton Murray, Masaaki Nagata, Martin Popel, Maja Popovi… — [Findings of the WMT24 General Machine Translation Shared Task: The LLM Era Is Here but MT Is Not Solved Yet](https://md.research.elamur.ai/free/en/papers/kocmi-2024-findings-wmt24-general-machine-translation/)
22. Cristóbal Parra Quesada and Manuela Cañizares Espada. 2024. Relationship of market capitalization of the ibex 35 to corporate social responsibility and transparency. Corporate Social Responsibility and Environmental Management, 31(4):3551–3572. [find](https://scholar.google.com/scholar?q=Crist%C3%B3bal+Parra+Quesada+and+Manuela+Ca%C3%B1izares+Espada%2E+2024%2E+Relationship+of+market+capitalization+of+the+ibex+35+to+corporate+social+responsibility+and+transparency%2E+Corporate+Social+Responsibility+an)
23. Sara Rajaee, Sebastian Vincent, Alexandre Berard, Marzieh Fadaee, Kelly Marchisio, and Tom Kocmi. 2026. Unlocking reasoning capability on machine translation in large language models. [find](https://scholar.google.com/scholar?q=Sara+Rajaee%2C+Sebastian+Vincent%2C+Alexandre+Berard%2C+Marzieh+Fadaee%2C+Kelly+Marchisio%2C+and+Tom+Kocmi%2E+2026%2E+Unlocking+reasoning+capability+on+machine+translation+in+large+language+models%2E)
24. Tom Kocmi, Eleftherios Avramidis, Rachel Bawden, Ondřej Bojar, Anton Dvorkovich, Christian Federmann, Mark Fishel, Markus Freitag, Thamme Gowda, Roman Grundkiewicz, Barry Haddow, Philipp Koehn, Benjamin Marie, Christof Monz, Makoto Morishita, Kenton Murray, Masaaki Nagata, Toshiaki Nakazawa, Martin… [find](https://scholar.google.com/scholar?q=Tom+Kocmi%2C+Eleftherios+Avramidis%2C+Rachel+Bawden%2C+Ond%C5%99ej+Bojar%2C+Anton+Dvorkovich%2C+Christian+Federmann%2C+Mark+Fishel%2C+Markus+Freitag%2C+Thamme+Gowda%2C+Roman+Grundkiewicz%2C+Barry+Haddow%2C+Philipp+Koehn%2C+Benjam)
25. Ricardo Rei, Jose Pombal, Nuno M. Guerreiro, João Alves, Pedro Henrique Martins, Patrick Fernandes, Helena Wu, Tania Vaz, Duarte Alves, Amin Farajian, Sweta Agrawal, Antonio Farinhas, José G. C. De Souza, and André Martins. 2024. Tower v2: Unbabel-IST 2024 submission for the general MT shared task.… [find](https://scholar.google.com/scholar?q=Ricardo+Rei%2C+Jose+Pombal%2C+Nuno+M%2E+Guerreiro%2C+Jo%C3%A3o+Alves%2C+Pedro+Henrique+Martins%2C+Patrick+Fernandes%2C+Helena+Wu%2C+Tania+Vaz%2C+Duarte+Alves%2C+Amin+Farajian%2C+Sweta+Agrawal%2C+Antonio+Farinhas%2C+Jos%C3%A9+G%2E+C%2E+De+So)
26. Aquia Richburg and Marine Carpuat. 2024. How multilingual are large language models fine-tuned for translation? [find](https://scholar.google.com/scholar?q=Aquia+Richburg+and+Marine+Carpuat%2E+2024%2E+How+multilingual+are+large+language+models+fine%2Dtuned+for+translation%3F)
27. Jörg Tiedemann and Yves Scherrer. 2017. [Neural machine translation with extended context](https://md.research.elamur.ai/free/en/papers/tiedemann-2017-neural-machine-translation-extended-context/). In Proceedings of the Third Workshop on Discourse in Machine Translation, pages 82–92, Copenhagen, Denmark. Association for Computational Linguistics.
28. Longyue Wang, Chenyang Lyu, Tianbo Ji, Zhirui Zhang, Dian Yu, Shuming Shi, and Zhaopeng Tu. 2023. [Document-level machine translation with large language models](https://md.research.elamur.ai/free/en/papers/wang-2023-document-level-machine-translation-large/).
29. Mike Zhang and Antonio Toral. 2019. [The effect of translationese in machine translation test sets](https://md.research.elamur.ai/free/en/papers/zhang-2019-effect-translationese-machine-translation-test/). In Proceedings of the Fourth Conference on Machine Translation (Volume 1: Research Papers), pages 73–81, Florence, Italy. Association for Computational Linguistics. 8.
30. Language Resource References Antonio Moreno-Sandoval, Yanco Amor Torterolo Orta, Sofía Micaela Roseti, Blanca Carbajo-Coronado, and Jordi Porta. 2025. Financial ES-EN parallel corpus from annual reports. [find](https://scholar.google.com/scholar?q=Language+Resource+References+Antonio+Moreno%2DSandoval%2C+Yanco+Amor+Torterolo+Orta%2C+Sof%C3%ADa+Micaela+Roseti%2C+Blanca+Carbajo%2DCoronado%2C+and+Jordi+Porta%2E+2025%2E+Financial+ES%2DEN+parallel+corpus+from+annual+repor)

*Updated 1 November 2026*