| language: | |
| - en | |
| - ja | |
| task_categories: | |
| - visual-question-answering | |
| - image-to-text | |
| pretty_name: Mecha Ja (English Translation) | |
| tags: | |
| - multimodal | |
| - translation | |
| - llm-jp-eval-mm | |
| # Mecha Ja (English Translation) | |
| This is an English translation of the [llm-jp/MECHA-ja](https://huggingface.co/datasets/llm-jp/MECHA-ja) dataset, | |
| translated using [plamo-translate](https://huggingface.co/pfnet/plamo-translate). | |
| ## Dataset Description | |
| This dataset is part of the [llm-jp-eval-mm](https://github.com/llm-jp/llm-jp-eval-mm) benchmark suite. | |
| The original Japanese questions and answers have been translated to English while preserving the visual content. | |
| ### Translation Details | |
| - **Translation Model**: pfnet/plamo-translate | |
| - **Fields Translated**: prompt, completion, instruction | |
| - **Original Language**: Japanese | |
| - **Target Language**: English | |
| ## Usage | |
| You can use this dataset with the llm-jp-eval-mm framework by specifying the language parameter: | |
| ```python | |
| from eval_mm.tasks import MechaJa | |
| # Use English version | |
| task = MechaJa(language="en") | |
| # Use original Japanese version | |
| task = MechaJa(language="ja") | |
| ``` |