Limits and Surprising Failures
Language models can struggle with basic arithmetic, such as multiplying large numbers. Despite being trained on a corpus, they may not have enough examples to accurately perform these calculations. However, there are ways to address these limitations and improve their performance.In this clip
From this podcast

Data Skeptic
Program Aided Language Models
Related Questions