Allow parallel runs for agents for n parameter

[MohammedYaseen97]

Hi, I was reading the SelfCheckGPT paper to allow the detection of halucination detection in answers by comparing different generations of the same chain, and noting the inconsistencies.

Wanted to try it out in a business application, but noticed that even if we choose n=5 in llm initialisation, the agent only generates one string... is it a good idea to include parallel answers to be generated by the agent run ?

If yes, I would like to go ahead and start work on it.

[Kirushikesh]

I was also looking on this, not only the agent but also extended to a chain, we can assess the factualness of a response from a chain using SelfCheckGPT methods.