Researcher, Peng Cheng Laboratory
1 paper at NeurIPS 2025
This study makes a key contribution by introducing a novel systematic framework to interpret the translation mechanisms of LLMs from a computational components perspective, an area previously unexplored.