I want to know which is the most effective way to structure my prompts which, apart from my question(s), need to incluyde some copy/pasted data from other sources. How best to signify which are the context blocks so the LLM, and more specifically ChatGPT, understands what is mine and what is copy/pasted? I'm also worried about cases where the provided context has words and language style which may look similar to my actual prompt. Examples follow.
1. Use triple backticks:
I'd like you to write this and that. Some guidance from the manual is:
```
info info data info info data data info ...
info info data info info data data info ...
info info data info info data data info ...
```
Also make sure take into account X, Y or Z based on:
info info data info info data data info ...
2. Use block quotes:
I'd like you to write this and that. Some guidance from the manual is:
> info info data info info data data info ...
> info info data info info data data info ...
> info info data info info data data info ...
Also make sure take into account X, Y or Z based on:
> info info data info info data data info ...
3. Use nothing to separate:
I'd like you to write this and that. Some guidance from the manual is:
info info data info info data data info ...
info info data info info data data info ...
info info data info info data data info ...
Also make sure take into account X, Y or Z based on:
info info data info info data data info ...
4. Use dashes to separate:
I'd like you to write this and that. Some guidance from the manual is:
---
info info data info info data data info ...
info info data info info data data info ...
info info data info info data data info ...
---
Also make sure take into account X, Y or Z based on:
info info data info info data data info ...
