Open lswainemoore opened 1 year ago
also
something is fishy!
Stepping back: I think this is a neat idea and well-intentioned, but I'm skeptical of the use of LLMs for grading all of these questions, many of which have a single correct answer, or a set of discrete correct answers that could be handled with some pretty simple pattern matching. I do think that for some of them, it could be an appropriate choice (e.g. "What did Susan B. Anthony do?"), but on the whole, I think if I was actually using this to study for the exam, the LLM's grading probably would've left me confused.
Other little issues I think caused by using LLMs:
I recognize that I'm sniping from the sidelines here, and I'm not exactly offering to make these changes myself ;), so I want to reiterate that I respect the intentions behind this project, and I think it has the potential to be a useful tool.
(This is after a bunch of JSON errors.)