How to use the trajectories of multiple points in the process of reasoning

Hello, all the examples in the inference stage of the code use the trajectory of a single point for inference. So how to use multiple points for inference? And how should the file format of the corresponding trajectory points be organized? In the paper, the maximum number of trajectory points is 16.

Start with the inference-stage examples referenced in the issue and compare their single-point trajectory input with the paper's maximum of 16 points. Determine how multiple trajectory points are represented and passed during inference, then document the expected file format and a multi-point usage example. Done means a newcomer can prepare the trajectory file and run multi-point inference without further clarification.

Written by the indexing model from the issue text.