DUKEAI/Alizee-OpenCodeReasoning-Phase3-1.4M
收藏Hugging Face2025-09-21 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/DUKEAI/Alizee-OpenCodeReasoning-Phase3-1.4M
下载链接
链接失效反馈官方服务：
资源简介：
Alizee OpenCodeReasoning Phase 3 Conformant 数据集是一个完全符合Phase 3规范的数据集，包含120万高质量Python代码示例，每个示例都有合成的提示和简洁的推理链。该数据集经过严格的处理，移除了所有格式化标签，通过代码分析生成合成提示，创建了简洁的推理要点，并正确区分了SFT和GRPO模式。此外，数据集通过AST哈希进行了去重，确保了所有Python代码的有效性。
The Alizee OpenCodeReasoning Phase 3 Conformant dataset is a fully conformant version of the Phase 3 dataset, containing 1.2 million high-quality Python code examples, each with synthetic prompts and concise reasoning chains. The dataset has been strictly processed to remove all formatting tags, generate synthetic prompts through code analysis, create concise bullet points for reasoning, and properly separate the SFT and GRPO schemas. Additionally, the dataset has been deduplicated using AST hashing and ensures the validity of all Python code.
提供机构：
DUKEAI
