How does recursive feature elimination work at a high level?

A. Add random features until training accuracy is perfect B. Eliminate observations instead of features C. Fit a model, remove the least important features, and repeat D. Replace all variables with one cluster label