Reinforcement learning works well on small models, costs under one dollar

Original: one of the biggest misconceptions about RL is that it's super expensive

Source: x.com ↗