Investigating the translation capabilities of Large Language Models trained on parallel data only

TL;DR

This paper introduces PLUME, a set of large language models trained solely on parallel data for Catalan, demonstrating competitive translation performance and providing insights into LLM translation capabilities without fine-tuning.

Contribution

The work presents the first LLMs trained exclusively on parallel data for translation, analyzing their performance and cross-lingual representations without instruction fine-tuning.

Findings

PLUME models perform comparably to encoder-decoder models on multiple translation tasks.

Prompt design significantly influences translation performance.

LLMs exhibit meaningful cross-lingual representations even without fine-tuning.

Abstract

In recent years, Large Language Models (LLMs) have demonstrated exceptional proficiency across a broad spectrum of Natural Language Processing (NLP) tasks, including Machine Translation. However, previous methods predominantly relied on iterative processes such as instruction fine-tuning or continual pre-training, leaving unexplored the challenges of training LLMs solely on parallel data. In this work, we introduce PLUME (Parallel Language Model), a collection of three 2B LLMs featuring varying vocabulary sizes (32k, 128k, and 256k) trained exclusively on Catalan-centric parallel examples. These models perform comparably to previous encoder-decoder architectures on 16 supervised translation directions and 56 zero-shot ones. Utilizing this set of models, we conduct a thorough investigation into the translation capabilities of LLMs, probing their performance, the impact of the different elements of the prompt, and their cross-lingual representation space.