Zhengxuan Wu

Assessing the accuracy of neuron explanations in language models reveals significant flaws.

Explore how interpretability illusions affect our view of neural networks.

A study assessing various methods for interpreting language model neurons.