We only have this paper's title and publication details.
Read it at the sourcePrevious work has shown that human evaluations in NLP are notoriously under-powered.
No takes yet. Share an insight, caveat, or question.
Howcroft et al. (2021) studied this question.