Hi all,
I know that LLMs are note state machine or similar but I don’t understand why at the step of trying to make a stand-alone question the LLM just don’t do it (see screenshot)
This is different from what seen during the lecture:
And of course the final answer is wrong and it’s the same we have at the end of L5…
The prompts seem to be the same, the only thing changing is the behaviour of the LLM, but I think this is really disappointing.
And in any case if it was a real situation: how we can debug it? Where the prompts are most into the template behind the scenes provided by LangChain.
What do you think?
Alexio