Hi everyone. I have built a chatbot that uses gpt4o and gpt-4o-mini. I provide the latest 4 of the interactions as a context to the chatbot. I have a usecase where I need to check for different if else cases and these cases can go upto 5 level deep. About 50% of the time; the bot fails to target the innermost cases and jumps to a conclusion. What might be the best areas I can put my effort into to tackle this issue?
p.s. since I am working on a chatbot; I would like the responses to be under 5s.
1 Like
Welcome to the forum!
I take it you do not have access to the o1 models.
I would not be surprised if the o1 models were able to do better with your need.
However the o1 models are often overkill for many problems, problems that are closer to NP-hard problems tend to be a better fit for using o1 model.
1 Like
but its mentioned that o1 models often trade speed for accuracy and critical thinking. And since I am working on a chatbot; I would also want some speed.
Ture.
In your initial question you should have noted that you would like the reply to be less than something like 5 seconds, then I would have not noted the 01 models.
Without more specifics one can only guess at what you are trying to do. As such I will pass on further replies.
Good luck with your problem.
1 Like