106 citations · 107 across the 2 of their papers we have counts for
Showing cs.CLShow all
2 papers · 1 filter
cs.CL2022★ 1 cited
Re-Examining Calibration: The Case of Question Answering
Chenglei Si, Chen Zhao, Sewon Min +1
For users to trust model predictions, they need to understand model outputs, particularly their confidence - calibration aims to adjust (calibrate) models' confidence to match expe…
cs.CL2021★ 106 cited
Between words and characters: A Brief History of Open-Vocabulary Modeling and Tokenization in NLP
Sabrina J. Mielke, Zaid Alyafeai, Elizabeth Salesky +8
What are the units of text that we want to model? From bytes to multi-word expressions, text can be analyzed and generated at many granularities. Until recently, most natural langu…