Asap7772/aime_backtracks_maxpav
收藏Hugging Face2025-02-01 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/Asap7772/aime_backtracks_maxpav
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了问题的提示、原始解决方案和步骤、正确性标志、数值序列、优势序列、回溯选择、优势最小值索引、值最小值索引、PAV最小值索引、优势最大值索引、值最大值索引、PAV最大值索引、最小值索引、PAV值序列、新解决方案、新正确性标志、当前响应、最佳响应标志、当前令牌数、总令牌数、唯一标识符、URL、目标答案、更新标志、数据索引和对话轮次等字段。训练集包含约26MB的字节和1260个示例。
The dataset includes fields for problem prompts, original solutions and steps, correctness flags, numeric sequences, advantage sequences, backtrack choices, indices and values for the minimum and maximum advantages, values, and PAV, new solutions, current responses, best response flags, current and total token counts, unique identifiers, URLs, target answers, update flags, data indices, and turn numbers. The training set contains approximately 26MB of bytes and 1260 examples.
提供机构：
Asap7772
