CommonsenseQA 2.0: Exposing the Limits of AI through Gamification
This work addresses the need for robust benchmarks in AI research to expose limitations in common sense reasoning, though it is incremental as it builds on existing benchmark construction methods.
The authors tackled the problem of creating challenging benchmarks for natural language understanding by introducing gamification to collect high-quality data, resulting in CommonsenseQA 2.0 with 14,343 yes/no questions where models like T5-based Unicorn (70.2% accuracy) and GPT-3 (52.9%) perform below human performance (94.1%).
Constructing benchmarks that test the abilities of modern natural language understanding models is difficult - pre-trained language models exploit artifacts in benchmarks to achieve human parity, but still fail on adversarial examples and make errors that demonstrate a lack of common sense. In this work, we propose gamification as a framework for data construction. The goal of players in the game is to compose questions that mislead a rival AI while using specific phrases for extra points. The game environment leads to enhanced user engagement and simultaneously gives the game designer control over the collected data, allowing us to collect high-quality data at scale. Using our method we create CommonsenseQA 2.0, which includes 14,343 yes/no questions, and demonstrate its difficulty for models that are orders-of-magnitude larger than the AI used in the game itself. Our best baseline, the T5-based Unicorn with 11B parameters achieves an accuracy of 70.2%, substantially higher than GPT-3 (52.9%) in a few-shot inference setup. Both score well below human performance which is at 94.1%.