Researchers compared performance of OpenAI’s GPT-3.5 and GPT-4.
It seems that the honeymoon phase for large language models (LLMs), introduced in the rush to make inroads in the generative AI space, is over. According to a study by researchers at Stanford and UC Berkeley, the performance of OpenAI’s LLMs has decreased significantly over time.
The researchers wanted to determine if these LLMs were improving, as they can be updated based on data, user feedback, and design changes.
Robert Way/iStock.
According to a study by researchers at Stanford and UC Berkeley, the performance of OpenAI’s LLMs has decreased significantly over time.
Leave a reply