Clarification on Data Source and Tool Invocation in V-ToolRL Agent Example

[zandfj]

In the paper, there is an example demonstrating the step-by-step visual reasoning process of the V-ToolRL agent. However, I cannot find the source of this example's data, as it is not present in the training or test datasets. Additionally, when testing your publicly available model, these two tools are rarely invoked. Could you please provide an explanation for this?