Visual Question Answering with Natural Language Explanations (VQA-NLE) is an important task that enhances the interpretability of artificial intelligent systems by generating answers to image-based questions accompanied by explanatory text. However, current VQA-NLE datasets are predominantly in English, limiting research in other languages, such as Vietnamese. To address this gap, we present ViVQA-X, a Vietnamese version of the VQA-X dataset, created through an automated translation pipeline that leverages multiple Large Language Models as evaluators to ensure high-quality translations. Our pipeline comprises translation, selection, and post-processing phases to identify the most accurate translations. We provide extensive analysis and benchmarking of multiple models to establish performance baselines for ViVQA-X. Our experiments reveal that NLX-GPT, a leading method on the original VQA-X dataset, achieves state-of-the-art performance on ViVQA-X. The source code and dataset are publicly available at https://github.com/duongtruongbinh/ViVQA-X .