AI safety & alignment

• Model organisms for emergent misalignmentarXiv, Jun 2025
• Convergent linear representations of emergent misalignmentarXiv, Jun 2025