Some questions about the Bayesian inverse planning part

[Jerry-hyl]

Hi Chuanyang,

I have some questions about the Bayesian inverse planning process. In this paper, it shows that we can compare the two goals or beliefs using this formular:

We can use LLM to get the policy function and accelerate the calculating process, but how can we get the item $P(b^t|b^{t-1},s^t)$? In the file BIP-ALM/inverse_symbolic_planner.py, I can only find the code of calculating the two policy items. Hope you can help clarify my question.

Thank you very much!