each specific comments in what follows

We would like to thank the reviewers for their valuable comments. Note that we also mentioned this in line 208. We will update our paper to reflect this comment. Re: the bandit version's approximation guarantee is much weaker than the offline version when the delays are not big, In this scenario, it is easy to see that online greedy becomes optimal. It is still true that the optimal solution of this instance is linked to the solution of the original 3-SA T. For the approximation ratio of the online greedy, it is true that we do not know whether our result is tight.