Lab take on a kids’ game reveals biases in ChatGPT

Abstract

Large language models behave similarly to humans in experiments where they must choose what information to keep or to discard. Large language models behave similarly to humans in experiments where they must choose what information to keep or to discard.