CoSpace: Benchmarking Continuous Space Perception Ability for Vision-Language Models

Figure 1. Illustration of continuous space perception: The left side depicts the construction of an image sequence representing a continuous space. The right side presents an example task related to the continuous space, showing how adjacent images are connected. Note that to answer the question correctly, one must recognize that the cars in images 1 and 2 are the same, as well as those in images 3 and 4.