3 ms·DeepMind: Directly Fine-Tuning Diffusion Models on Differentiable Rewards2 points by tim_sw 3y ago