How outdated information hides in LLM token generation probabilities
(anj.ai)
The internet usually has the correct answer somewhere, but it’s also full of conflicting and outdated information. How do large language models (LLMs) such as ChatGPT, trained on internet scale data, handle cases where there’s conflicting or outdated information? (Hint: it’s not always the most recent answer as of the knowledge cutoff date; think about what LLMs are trained to do)
The internet usually has the correct answer somewhere, but it’s also full of conflicting and outdated information. How do large language models (LLMs) such as ChatGPT, trained on internet scale data, handle cases where there’s conflicting or outdated information? (Hint: it’s not always the most recent answer as of the knowledge cutoff date; think about what LLMs are trained to do)