Results 4 comments of Leon

![image](https://github.com/user-attachments/assets/5c54b50a-5f6b-42e6-af22-0bffee15bc41) 另外,原代码中这里少乘了一个 $\phi(s_t)$

- [x] rebased onto latest master - [x] Demo in attachment [prompt_set_demo.zip](https://github.com/user-attachments/files/20442806/prompt_set_demo.zip) @hsluoyz

fixed @hsluoyz https://github.com/user-attachments/assets/1b0235bc-f674-4042-8688-cd82b07910c0