Comment by bugglebeetle

Glad to see the NER-like task performed the best, as I was just about to test something like this for comparison with a fine-tuned BERT model. Any idea about the training costs for this task?