Research Reveals LLMs Can Learn to Resist Reinforcement Learning Through Exploration Hacking

Friday, May 1, 2026