In manga translation, visual context outside speech-bubble text, including speaker identities, referents, and scene tone, is depicted in page images, making it natural to refer to images during translation evaluation. However, most conventional manga translation evaluations rely solely on text-only automatic metrics, leaving the impact of image input insufficiently explored. We propose a practical validation framework that employs the same vision-language model (VLM) as both a text-only judge (TOJ) and an image-aware judge (IAJ) to investigate evaluation divergences associated with image-aware evaluation procedures. We use a Japanese-English manga dataset of official translations aligned at the page and speech-bubble levels. First, adding page images to TOJ enabled the primary VLM to detect more degraded translations. Second, we analyzed evaluation divergences between TOJ and IAJ, finding different error counts and error-category distributions. Human verification did not reveal a consistent preference between TOJ and IAJ.

We participated in the constrained English–Japanese track of the WMT 2025 General Machine Translation Task.Our system collected the outputs produced by multiple subsystems, each of which consisted of LLM-based translation and reranking models configured differently (e.g., prompting strategies and context sizes), and reranked those outputs.Each subsystem generated multiple segment-level candidates and iteratively selected the most probable one to construct the document translation.We then reranked the document-level outputs from all subsystems to obtain the final translation.For reranking, we adopted a text-based LLM reranking approach with a reasoning model to take long contexts into account.Additionally, we built a bilingual dictionary on the fly from parallel corpora to make the system more robust to rare words.

We participated in the constrained track for English-Japanese and Japanese-Chinese translations at the WMT 2024 General Machine Translation Task. Our approach was to generate a large number of sentence-level translation candidates and select the most probable translation using minimum Bayes risk (MBR) decoding and document-level large language model (LLM) re-ranking. We first generated hundreds of translation candidates from multiple translation models and retained the top 30 candidates using MBR decoding. In addition, we continually pre-trained LLMs on the target language corpora to leverage document-level information. We utilized LLMs to select the most probable sentence sequentially in context from the beginning of the document.