Summary of How Multilingual Are Large Language Models Fine-tuned For Translation?, by Aquia Richburg and Marine Carpuat
How Multilingual Are Large Language Models Fine-Tuned for Translation?
by Aquia Richburg, Marine Carpuat
First submitted to arxiv on: 30 May 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The new paradigm of fine-tuning large language models (LLMs) on parallel text has outperformed dedicated translation systems in machine translation tasks. This study investigates whether this approach can be scaled up for massively multilingual machine translation or if it requires fine-tuning separate models for specific language pairs. The authors evaluate the translation quality of the TOWER family of LLMs on 132 translation tasks from the FLORES-200 data, finding that fine-tuning improves translation quality even for zero-shot languages, but with varying impact depending on the language pairs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study takes a new approach to machine translation by fine-tuning large language models (LLMs) on parallel text. Researchers are trying to figure out if this method can be used for lots of different languages or if it only works well for a few specific ones. They tested this approach on many different translation tasks and found that it actually makes the translations better, even when working with languages they haven’t seen before. But the results were a bit mixed, depending on which languages were being translated. |
Keywords
» Artificial intelligence » Fine tuning » Translation » Zero shot
