Fine tuning the BERT LLM for essay grading.

I take a neural network based ordinal regression approach to the problem, and add a classifier at the end of the LLM to fintune.

The metric is Quadratic Weighted Kappa and I achieved a score of 0.78 on the test set.