Akash comments on Measuring and Improving the Faithfulness of Model-Generated Reasoning