ArtWhisperer

ArtWhisperer是一个由斯坦福大学创建的数据集，旨在研究人与AI在艺术创作中的交互。该数据集通过一个在线游戏收集，用户需迭代找到能生成与目标图像相似的AI图像的提示。数据集包含51,026次交互，涉及2,250名用户和191个独特的目标图像。创建过程涉及用户与AI的多次交互，以找到最佳提示。该数据集主要用于研究人与AI的交互行为，评估AI的可操控性，并展示了ArtWhisperer数据集的通用实用性。

ArtWhisperer is a dataset developed by Stanford University to study human-AI interaction in artistic creation. Collected via an online game, it requires users to iteratively find prompts that can generate AI images similar to the target ones. The dataset contains 51,026 interaction records, involving 2,250 users and 191 unique target images. Its creation involves multiple rounds of interactions between users and AI to identify the optimal prompts. This dataset is primarily used to research human-AI interaction behaviors, evaluate the controllability of AI, and demonstrate the general applicability of the ArtWhisperer dataset.