1 paper
Mingyu Jin, Yutong Yin, Jingcheng Niu +7
In this work, we investigate how Large Language Models (LLMs) adapt their internal representations when encountering inputs of increasing difficulty, quantified as the degree of ou…