We thank the reviewers for their effort and for their thoughtful comments.  We will integrate all of them while preparing the final version. We are glad to hear that all reviewers have enjoyed reading the paper.$
Reviewer 1
========

Many thanks for the careful reading and the detailed comments---we will implement all of them in our final version. 

* You are correct that the key to getting improved regret is really the (statistical) independence of the losses from the feedback graphs, rather than the stochasticity of the losses.  We thought that this key point should be discussed only after the formal presentation of the model and the main results.  However, in light of your comment we will definitely consider discussing this earlier in the paper.