Measuring the Faithfulness of Thinking Drafts in Large Reasoning Models | Read Paper on Bytez