This is a borderline paper that discusses the problem of robustness with two sources. The reviewers thought the approach was novel; however, there is a lack of theoretical analysis and question about the method's generalization to more complex scenarios.