We consider the detection of multiple outliers in Exponential and Pareto samples -- as well as general samples that have approximately Exponential or Pareto tails, thanks to Extreme Value Theory. It is shown that a simple "robust'' modification of common test statistics makes inward sequential...