Predicting Fine-Tuning Performance with Probing
Large NLP models have recently shown impressive performance in language understanding tasks, typically evaluated by their fine-tuned performance. Alternatively, probing has received increasing...
arxiv.org