Explainability / Model Interrogation?

[zkghost]

Have you thought at all about incorporating something like this into the project?

Maybe I just need to take a deeper look at the model being used but being able to see why certain questions/answers are being generated would help me verify that the output is sound.