
Less Sycophancy, Stronger Refusal? Lessons for AI Safety from Mechanistic Interpretability
Research
New papers, breakthroughs and academic AI research

Research

Research

Research

Research

Research

Research

Research

Research

Research

Research

Research

Research