inclusionAI/Ring-lite-rl-data

Ring-lite-rl-data数据集是一个专门为强化学习设计的数学和编程领域高质量问题子集。该数据集包括超过3.9万个经过严格筛选的数学问题，来源于开源数据集和数学竞赛集合，以及大约8400个经过验证的编程问题，来源于编程竞赛资源和开源编程竞赛数据。所有问题都包括经过验证的“接受”解决方案和测试用例。数据集经过严格的三阶段筛选流程，包括数据清洗、答案验证和数据标注。数学领域的字段包括问题陈述、正确答案、问题类型和唯一问题ID；编程领域的字段包括详细的编程问题描述、正确Python解决方案代码、实现语言、问题类型、测试用例列表、源数据集、编程语言、难度评分和唯一问题ID。

The Ring-lite-rl-data dataset is a curated subset of high-quality problems in mathematics and code domains designed for reinforcement learning. This dataset includes over 39,000 rigorously curated math problems sourced from open datasets and math contest collections, as well as approximately 8,400 verified coding problems from programming competition resources. All problems come with validated Accepted solutions and test cases. The dataset has undergone a rigorous three-stage curation process including data cleansing, answer verification, and data annotation. Fields in the mathematics domain include the problem statement, correct answer, problem category, and unique problem ID, while the code domain fields include detailed programming problem descriptions, correct Python solution code, implementation language, problem category, a list of validated test cases, source dataset, programming language, difficulty score, and unique problem ID.