mlfoundations-dev/MATH500_evalchemy

该数据集包含两个配置：默认配置（default）和实例配置（instances）。每个配置都有上下文（context）和生成参数（gen_kwargs）。上下文包含内容和角色两个字段，生成参数包含是否采样、最大新标记数、种子和温度等字段。默认配置还包括重复索引和请求索引字段，实例配置包括包含请求ID的元数据字段。数据集的训练分片包含500个示例，分别对应两种配置。

The dataset consists of two configurations: default and instances. Each configuration has context and gen_kwargs. The context includes content and role fields, while gen_kwargs includes fields such as do_sample, max_new_tokens, seed, and temperature. The default configuration also includes repeat_idx and request_idx fields, and the instances configuration includes metadata with a request_id field. The training split of the dataset contains 500 examples for each configuration.