Tall Tales at Different Scales: Evaluating Scaling Trends For Deception In Language Models — AI Alignment Forum