##article.return## Investigating Training and Generalization in Faithful Self-Explanations of Large Language Models Download Download PDF