Paper Reasoning Models Don't Always Say What They Think Anthropic · 2025 · arXiv:2505.05410 Read the paper → chain-of-thoughtfaithfulnessdeception
