This paper designs a network which can map the multi-channel audio from an array to the audio of a source at desired angular region. There has been active discussions after the author rebuttal. In the end, all reviewers agree that there is sufficient novelty in the methodology. There have also been minor concerns on the experimental setup being less realistic, not demonstrating whether the method is robust under more challenging scenarios. But reviewers also reached the consensus that the current result is significant enough for a conference paper. We hope the authors take advantage of the insightful reviews in future work.