1 paper
Parker Riley, Daniel Deutsch, George Foster +3
Reliable human evaluation is critical to the development of successful natural language generation models, but achieving it is notoriously difficult. Stability is a crucial require…