AI Models Sometimes Act Against Their Own Stated Moral Judgment

Original: Principled Under Pressure: Post-Training Decides Whether LLMs Act on Their Own Moral Judgment

Source: arxiv.org ↗