# Translation Asymmetry in LLMs as a Data Augmentation Factor: A Case Study for 6 Romansh Language Varieties

arXiv: https://arxiv.org/abs/2603.25489 · HTML: https://mlsanity.com/paper/2603.25489/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 41 references that are in the map.
- First posted on arXiv 26 Mar 2026.

## Builds on (most cited first, 22 of 22)

- [A Call for Clarity in Reporting BLEU Scores](https://mlsanity.com/paper/1804.08771/) (2018)
- [Improving Neural Machine Translation Models with Monolingual Data](https://mlsanity.com/paper/1511.06709/) (2015)
- [xCOMET: Transparent Machine Translation Evaluation through Fine-grained Error Detection](https://mlsanity.com/paper/2310.10482/) (2023)
- [MADLAD-400: A Multilingual And Document-Level Large Audited Dataset](https://mlsanity.com/paper/2309.04662/) (2023)
- [FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language](https://mlsanity.com/paper/2506.20920/) (2025)
- [WMT24++: Expanding the Language Coverage of WMT24 to 55 Languages & Dialects](https://mlsanity.com/paper/2502.12404/) (2025)
- [Dictionary-based Phrase-level Prompting of Large Language Models for Machine Translation](https://mlsanity.com/paper/2302.07856/) (2023)
- [Domain Robustness in Neural Machine Translation](https://mlsanity.com/paper/1911.03109/) (2019)
- [Investigating the Translation Performance of a Large Multilingual Language Model: the Case of BLOOM](https://mlsanity.com/paper/2303.01911/) (2023)
- [Segment Any Text: A Universal Approach for Robust, Efficient and Adaptable Sentence Segmentation](https://mlsanity.com/paper/2406.16678/) (2024)
- [Domain, Translationese and Noise in Synthetic Data for Neural Machine Translation](https://mlsanity.com/paper/1911.03362/) (2019)
- [BOUQuET: dataset, Benchmark and Open initiative for Universal Quality Evaluation in Translation](https://mlsanity.com/paper/2502.04314/) (2025)
- [Using Monolingual Data in Neural Machine Translation: a Systematic Study](https://mlsanity.com/paper/1903.11437/) (2019)
- [Shortcomings of LLMs for Low-Resource Translation: Retrieval and Understanding are Both the Problem](https://mlsanity.com/paper/2406.15625/) (2024)
- [Causal Estimation of Tokenisation Bias](https://mlsanity.com/paper/2506.03149/) (2025)
- [Scaling Low-Resource MT via Synthetic Data Generation with LLMs](https://mlsanity.com/paper/2505.14423/) (2025)
- [SwiLTra-Bench: The Swiss Legal Translation Benchmark](https://mlsanity.com/paper/2503.01372/) (2025)
- [Terminology-Aware Translation with Constrained Decoding and Large Language Model Prompting](https://mlsanity.com/paper/2310.05824/) (2023)
- [Rule-Based, Neural and LLM Back-Translation: Comparative Insights from a Variant of Ladin](https://mlsanity.com/paper/2407.08819/) (2024)
- [Expanding the WMT24++ Benchmark with Rumantsch Grischun, Sursilvan, Sutsilvan, Surmiran, Puter, and Vallader](https://mlsanity.com/paper/2509.03148/) (2025)
- [The Mediomatix Corpus: Parallel Data for Romansh Language Varieties via Comparable Schoolbooks](https://mlsanity.com/paper/2508.16371/) (2025)
- [Robust Language Identification for Romansh Varieties](https://mlsanity.com/paper/2603.15969/) (2026)

## Built on by (most cited first, 2 of 2)

- [Reinforcement Learning Elicits Contextual Learning of Unseen Language Translation](https://mlsanity.com/paper/2606.06428/) (2026)
- [Back-Translation Still Matters: Leveraging Monolingual Data for Low-Resource Machine Translation with LLMs](https://mlsanity.com/paper/2610.18818/) (2026)

## Abstract (arXiv)

Recent strategies for low-resource machine translation rely on LLMs to generate synthetic data based on text in higher-resource languages. We revisit this idea for Romansh, a language with 6 distinct varieties. LLMs tend to confuse these varieties when translating into Romansh, but they are quite good at translating out of Romansh into a high-resource language such as German. Due to this asymmetry, the direction of data augmentation is a crucial choice. We find that contrary to recent strategies, creating synthetic translations into the higher-resource language is the superior approach, and only this approach allows us to surpass a Gemini 3 Pro baseline on German-Romansh translation (+23 BLEU over Gemini in the lowest-resource variety). A human evaluation confirms that our experiments yield the first model that generates fluent translations in the individual Romansh varieties.