# Natural language understanding

читать 2 мин.
0 29

The article defines general linguistic intelligence and evaluates state-of-the-art natural language understanding models against this definition. The authors propose a new evaluation metric and find that while models have made progress in generalizing to many tasks, they still require a lot of in-domain training and are prone to overfitting. The article discusses missing components and suggests ways to make progress toward general linguistic intelligence.

Перепост
Прочитать полностью