Benchmark dataset for safety assessment under blast loading

This dataset provides a benchmark comprising 50 curated assessment tasks designed to evaluate the performance of large language model (LLM)-based multi-agent frameworks. Each item represents an independent task with both simple and complex task descriptions, along with corresponding ground-truth references.本数据集为一款包含50个精心甄选评估任务的基准测试集，旨在评估基于大语言模型（Large Language Model，LLM）的多智能体框架的性能。每项任务均为独立任务，同时配有简易与复杂两种任务描述，以及对应的真值参考。