Nathan Helm-Burger comments on Sparse trinary weighted RNNs as a path to better language model interpretability

[Nathan Helm-Burger]

Thanks! And I’m excited to hear more about your work. It sounds like if it did work, the results would be quite interesting.