DopeorNope/FFT-naive-50k-minif2f
收藏资源简介:
该数据集包含244个训练样本,专注于自然语言到形式化语言的生成任务,可能涉及数学定理证明,特别是与Lean定理证明器相关。特征包括模型、数据集、分割、名称、声明名称使用、样本索引、温度、top_p、top_k、最大令牌数、启用思考、去除注释标记、用户提示、非正式内容、参考正式声明、原始生成、Lean4生成和Lean文件生成,用于训练或评估AI模型在将非正式数学陈述转换为形式化Lean代码方面的性能。
This dataset contains 244 training examples focused on natural language to formal language generation tasks, likely related to mathematical theorem proving, particularly with the Lean proof assistant. Features include model, dataset, split, name, decl_name_used, sample_idx, temperature, top_p, top_k, max_tokens, enable_thinking, strip_comment_markers, prompt_user, informal, ref_formal_statement, gen_raw, gen_lean4, and gen_lean_file, designed for training or evaluating AI models in converting informal mathematical statements into formal Lean code.




