I keep trying to get this to work
It always gets the columns the wrong way round, i.e. the most negative score is treated as most positive. It works fine to tell it that it is wrong, but I want to use the api.
Unfortunately, the answer is that this is the wrong tool for the job.
However, given the new code interpreter, you could probably get GPT to generate the code necessary to evaluate the data in the way you are seeking. The code interpreter would then run the calculations.
But as is, GPT is known to lack in it’s numerical abilities.
2 Likes
Cheers, I can easily send the average score instead, but weird how it gets it wrong with every model.
1 Like