SuperGLUE is a benchmark for AI developed to outline the research progress on different sets of language tasks. It has been proposed by researchers at DeepMind, Facebook, University of Washington, and New York University. This new benchmark was developed on the already existing GLUE benchmark.
When the benchmark was initially introduced, there was a significant gap between the top-performing model and human performance. Recently, T5+Meena, developed by Google, and DeBERTa developed by Microsoft, have become the first AI models to surpass human baselines on the leaderboard.