sensenova/SenseNova-SI-800K

SenseNova-SI-800K是一个旨在提升多模态基础模型空间智能的数据集，属于SenseNova-SI系列的一部分。该数据集包含80万个多样化的数据样本，采用JSONL格式存储，每个样本包含唯一的ID、对话记录（人类与GPT之间的对话轮次）和图像路径。数据集支持视觉问答和问答任务，旨在通过系统化的数据构建方法，提升模型在广泛空间智能基准测试中的表现，同时保持强大的通用多模态理解能力。README还提到了基于该数据集训练的模型，展示了其在空间智能基准测试中的显著改进和竞争力。

SenseNova-SI-800K is a dataset designed to enhance spatial intelligence in multimodal foundation models, part of the SenseNova-SI family. It comprises 800,000 diverse data samples stored in JSONL format, each containing a unique ID, conversations (dialogue turns between humans and GPT), and image paths. The dataset supports visual-question-answering and question-answering tasks, aiming to improve model performance across a broad range of spatial intelligence benchmarks while maintaining strong general multimodal understanding capabilities through a principled approach to data curation. The README also highlights a model trained on this dataset, demonstrating significant improvements and competitive performance in spatial intelligence benchmarks.