Almita

该数据集名为Almita，旨在评估人工智能代理在处理复杂对话任务时的表现，重点关注行动回忆和目标完成度这两个指标。该数据集包含了一系列复杂的使用场景，评估指标不仅关注单个对话轮次，也关注端到端的任务完成情况。其规模涉及与多个使用场景的复杂交互，任务则是评估人工智能代理在面向任务的对话中的性能。

This dataset, termed Almita, is developed to evaluate the performance of AI Agents in complex conversational tasks, with two primary metrics: action recall and goal achievement. It encompasses a variety of complex real-world usage scenarios, and its evaluation metrics cover not only individual conversation turns but also end-to-end task completion results. In terms of scale, it involves intricate interactions across multiple usage scenarios, with the core evaluation task being to assess the performance of AI Agents in task-oriented dialogues.