ShengbinYue/Long-short-Trajectory

该轨迹数据集是为了优化多智能体框架中的长短轨迹学习而构建的，包含了从12个以上的知识密集型任务中收集的样本，能够覆盖各种指令语义和格式。数据集分为长轨迹数据集和短轨迹数据集两种类型，适用于长短轨迹学习的不同阶段，使得框架能够结合任务特定训练和端到端推理的优势。

The Trajectory dataset is constructed to optimize long-short trajectory learning in our multi-agent framework, containing samples collected from over 12 knowledge-intensive tasks, covering a variety of instruction semantics and formats. The dataset consists of two types: Long-trajectory Datasets and Short-trajectory Datasets, which are applied in different stages of long short-trajectory learning, allowing the framework to combine the benefits of task-specific training with the power of end-to-end reasoning.