Theory of Mind Question Evaluation Benchmark
收藏资源简介:
本研究构建了一个开放式的理论思维评估基准,名为Theory of Mind Question Evaluation Benchmark,包含30个基于意图类别的图像场景。这些图像从多个平台获取,每个图像都经过严格筛选,确保能够清晰传达意图,并伴有详细描述、视觉线索和未来推理的标注。该数据集旨在评估视觉语言模型在理解人类意图方面的能力,涉及情感和动作两种意图类别,适用于多模态理解和社交智能的研究。
This study constructs an open Theory of Mind evaluation benchmark named Theory of Mind Question Evaluation Benchmark, which includes 30 image scenarios categorized by intent types. These images are sourced from multiple platforms, and each has undergone rigorous screening to ensure it clearly conveys intent, with annotations of detailed descriptions, visual cues and future reasoning. This dataset aims to evaluate the capability of vision-language models in understanding human intentions, covering two intent categories: emotional and behavioral intentions, and is suitable for research on multimodal understanding and social intelligence.




