tanhuajie2001/Reason-RFT-CoT-Dataset

Reason-RFT CoT数据集是为了支持Reason-RFT视觉推理强化训练框架而构建的。该数据集包含多种视觉推理任务，如视觉计数、结构感知和空间转换等，旨在评估视觉认知、几何理解和空间泛化能力。数据集分为训练集和测试集，并提供了详细的Chain-of-Thought (CoT)注释，以帮助视觉语言模型(VLMs)激活推理潜力。

The Reason-RFT CoT Dataset is constructed to support the Reason-RFT reinforcement fine-tuning framework for visual reasoning. The dataset encompasses various visual reasoning tasks such as visual counting, structure perception, and spatial transformation, aiming to assess visual cognition, geometric understanding, and spatial generalization capabilities. The dataset is divided into training and test sets and is provided with detailed Chain-of-Thought (CoT) annotations to help Vision-Language Models (VLMs) activate their reasoning potential.