sensenova/SenseNova-MARS-Data

该数据集提供了Wikipedia语料库以及完全处理的训练和评估数据，可直接用于SenseNova-MARS项目。为确保训练和评估脚本正常运行，必须按照指定的目录结构下载并放置这些资源。数据集包含预处理的Wikipedia知识库（包含Faiss索引文件和格式化文本）、专门为SenseNova-MARS训练管道格式化的训练数据集以及用于视觉推理和工具使用任务的评估数据。

This dataset provides the Wikipedia corpus and fully processed training and evaluation data that can be used out of the box for SenseNova-MARS. To ensure the training and evaluation scripts function correctly, these resources must be downloaded and placed into your code directory following the structure below. The dataset consists of a pre-processed knowledge base containing Faiss index files and formatted text for high-speed local retrieval, processed training datasets specifically formatted for the SenseNova-MARS training pipeline, and processed evaluation data used to measure model performance on visual reasoning and tool-use tasks.