LLM Sensitivity Issues
François discusses the extreme sensitivity of LLMs to phrasing changes, highlighting how minor adjustments can significantly impact performance. He contrasts the notion of prompt engineering as a solution with the idea that LLMs often rely on superficial pattern matching rather than true understanding. Additionally, he notes that while LLMs can solve familiar problems, they struggle with unfamiliar tasks, revealing limitations in their adaptability and comprehension.In this clip
From this podcast
Machine Learning Street Talk (MLST)
It's Not About Scale, It's About Abstraction - Francois Chollet
Related Questions