VGS-AI/OpenR1-VM

这个数据集包含了250万个用于训练价值模型的推理轨迹，来源于论文《Value-Guided Search for Efficient Chain-of-Thought Reasoning》。每个问题包含56个来自DeepSeek-R1-Distill-Qwen模型（1.5B, 7B, 14B, 32B）的Chain-of-Thoughts（CoTs）。该数据集用于提高长上下文推理模型的效率。

This dataset contains 2.5 million reasoning traces used for training a value model in the paper Value-Guided Search for Efficient Chain-of-Thought Reasoning. For each problem, it includes 56 Chain-of-Thoughts (CoTs) sampled from DeepSeek-R1-Distill-Qwen models (1.5B, 7B, 14B, 32B). This data is used to improve the efficiency of long-context reasoning models.