Reviews: Hindsight Credit Assignment

Neural Information Processing Systems 

Having read through the other reviews and the author response I will maintain my review of a 6. I really like the core idea of the paper and would be happy if it were accepted based on that alone. I appreciate the author's clarification of the experiments, and I now have a much clearer understanding of what was done. With that said I was a little disappointed that the answer to my question about bootstrapping was basically "a quirk of the learning dynamics". In general, the main reason I have not raised my score is that I found the significance of the experiments to be hard to judge and they don't necessarily clearly illustrate the merit of the approach.