Superglue: A Stickier Benchmark for General-Purpose Language Understanding Systems

By Alex Wang et al

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Related Work
3. Superglue Overview
3.1 Design Process
3.2 Selected Tasks

Summary

The paper introduces Superglue, a new benchmark for language understanding tasks that aims to provide a more rigorous test compared to the existing GLUE benchmark. It includes more challenging tasks, diverse task formats, comprehensive human baselines, improved code support, and refined usage rules. The paper discusses the design process behind Superglue and details the selected tasks included in the benchmark.