4 ms·Scalable agent alignment via reward modeling - DeepMind Safety Research16 points by seriousssam 8y ago