1 code implementation • 14 Feb 2023 • Ard Snijders, Douwe Kiela, Katerina Margatina
We show that four popular active learning schemes fail to outperform random selection when applied to unlabelled pools comprised of multiple data sources on the task of natural language inference.
1 code implementation • RC 2020 • Andrew Harrison, Rahel Habacker, Ard Snijders, Mathias Parisot
Across multiple datasets, the authors show that with the impermissible tokens removed the model accuracy drops, implying their usage in prediction.