hamishivi/agent-task-swe-smith

该数据集包含59,136个训练示例，每个示例包括多轮对话消息（含内容和角色）、真实答案、数据集来源、环境配置（如环境名称、镜像和任务ID）以及源信息。数据用于支持对话系统或任务执行模型的训练，涉及结构化环境交互和答案验证。

This dataset contains 59,136 training examples, each comprising multi-turn dialogue messages (with content and role), ground truth answers, dataset source, environment configuration (including environment name, image, and task ID), and source information. It is designed to support training for dialogue systems or task-execution models, involving structured environment interactions and answer validation.