Covert underperformance on safety research tasks · OpenAI Alignment

Research on aligning AI with human values and intent, and reports documenting model failures.

saved by

• Tazik Sh
• Bhagyesh Kumar
• Linus Tang