Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitors Paper • 2512.11949 • Published Dec 12, 2025 • 1