HI-TOM
收藏资源简介:
HI-TOM数据集是由密歇根大学创建的,用于评估大型语言模型在高级心智理论(ToM)推理能力。该数据集包含300个故事,每个故事涉及多个代理之间的交互,并提出从零阶到四阶的ToM推理问题。数据集通过自动生成故事并手动检查质量,确保了内容的多样性和挑战性。HI-TOM数据集的应用领域包括复杂场景下的有效沟通、情感支持和同理心交流,旨在解决当前大型语言模型在高级ToM推理方面的局限性。
The HI-TOM Dataset, developed by the University of Michigan, is designed to evaluate the advanced Theory of Mind (ToM) reasoning capabilities of Large Language Models (LLMs). This dataset includes 300 stories, each involving interactions between multiple agents, and poses ToM reasoning questions ranging from zero-order to fourth-order levels. The dataset ensures diverse and challenging content through automatic story generation and manual quality inspection. The HI-TOM Dataset has applications in effective communication, emotional support, and empathic interaction in complex scenarios, and aims to address the current limitations of large language models in advanced ToM reasoning.

- 1HI-TOM: A Benchmark for Evaluating Higher-Order Theory of Mind Reasoning in Large Language Models密歇根大学 · 2023年



