PrimeIntellect/INTELLECT-3-RL

该数据集包含四个配置：code、logic、math和science，每个配置代表一个子集。code子集包含8579个训练示例，特征包括info和question，以及avg@8_qwen3_4b_instruct_2507评分；logic子集包含11647个训练示例，特征包括question、info和avg@16_qwen3_4b_instruct_2507评分；math子集包含21161个训练示例，特征包括question、answer和多个评分（如avg@8_qwen3_4b_thinking_2507）；science子集包含29307个训练示例，特征包括question、answer、info（包含domain字段）和avg@16_qwen3_4b_instruct_2507评分。整体数据集专注于代码、逻辑、数学和科学领域的问题，可能用于评估或训练AI模型。

This dataset consists of four configurations: code, logic, math, and science, each representing a subset. The code subset includes 8579 training examples with features such as info, question, and avg@8_qwen3_4b_instruct_2507 score; the logic subset includes 11647 training examples with features like question, info, and avg@16_qwen3_4b_instruct_2507 score; the math subset includes 21161 training examples with features including question, answer, and multiple scores (e.g., avg@8_qwen3_4b_thinking_2507); the science subset includes 29307 training examples with features such as question, answer, info (containing a domain field), and avg@16_qwen3_4b_instruct_2507 score. The overall dataset focuses on questions in the domains of code, logic, mathematics, and science, likely used for evaluating or training AI models.