Exploring the fundamentals of language mapping, a simple synthetic language with a limited vocabulary was tested to understand model performance. While transformers achieved 100% accuracy effortlessly, the state space model struggled to solve these basic language tasks, revealing intriguing differences in their capabilities.
