est-ai/math-reasoning-dpo
收藏Hugging Face2025-08-06 更新2025-11-01 收录
下载链接：
https://hf-mirror.com/datasets/est-ai/math-reasoning-dpo
下载链接
链接失效反馈官方服务：
资源简介：
数学推理DPO数据集包含数学推理问题及其选择和拒绝的响应，用于直接偏好优化(DPO)和语言模型的偏好学习。数据集遵循ShareGPT格式，包含对话轮次、首选响应和较不首选的响应。首选响应提供详细的推理和正确的数学解答，而较不首选的响应可能包含错误或不完整的推理。
This dataset contains mathematical reasoning problems with chosen and rejected responses, designed for Direct Preference Optimization (DPO) and preference learning of language models. The dataset follows the ShareGPT format, including conversation turns, preferred response with detailed reasoning and correct solution, and less preferred response which may contain errors or incomplete reasoning.
提供机构：
est-ai
