weblab-llm-competition-2025-bridge/difficult_problem_dataset_v4_500

这是一个基于SDG框架生成的，包含500个问题-答案对的数据集，适用于推理模型训练。每个问题都会经过自动生成、质量保证和多样性筛选、进化方法增强、推理过程自动生成等步骤。数据集中的每个实例都包括问题、推理和答案、使用模型、学科领域、数据来源、评分以及交互消息。

This is a dataset containing 500 question-answer pairs generated using the SDG framework, suitable for training reasoning models. Each question goes through steps such as automatic generation, quality assurance and diversity filtering, enhancement through evolutionary methods, and automatic generation of reasoning processes. Each instance in the dataset includes the question, reasoning and answer, used model, academic field, data source, score, and interaction messages.