Salesforce/HERB

HERB是一个评估大型语言模型代理在深度搜索和长上下文推理方面的能力的基准数据集。该数据集通过模拟企业工作流程中产品规划、开发和支持的各个阶段，生成具有真实噪声和保证有真实答案的多跳问题的互联内容。

HERB is a benchmark for evaluating the ability of large language model agents to perform deep search and long context reasoning. The dataset is generated by simulating the stages of product planning, development, and support in enterprise workflows, producing interconnected content with realistic noise and multi-hop questions with guaranteed ground-truth answers.