感觉Discriminator【总之】受ground-truth限制。
另外,
感觉Discriminator如果好了,强化学习的Reward就有了,Reward就Enough了(https://deepmind.com/research/publications/2021/Reward-is-Enough)
感觉Discriminator【总之】受ground-truth限制。
另外,
感觉Discriminator如果好了,强化学习的Reward就有了,Reward就Enough了(https://deepmind.com/research/publications/2021/Reward-is-Enough)