In2x at WMT25 Translation Task

TL;DR

This paper describes In2x's submission to the WMT25 translation task, focusing on extending large language models to Japanese and low-resource languages through novel data and reward model strategies.

Contribution

It introduces a generalizable paradigm for adapting large language models to low-resource languages, emphasizing data construction and reward modeling.

Findings

Achieved competitive translation performance on Japanese tasks.

Developed new data construction methods for low-resource language translation.

Proposed a reward model framework for improving translation quality.

Abstract

This paper presents the open-system submission by the In2x research team for the WMT25 General Machine Translation Shared Task. Our submission focuses on Japanese-related translation tasks, aiming to explore a generalizable paradigm for extending large language models (LLMs) to other languages. This paradigm encompasses aspects such as data construction methods and reward model design. The ultimate goal is to enable large language model systems to achieve exceptional performance in low-resource or less commonly spoken languages.