@javidiqbal

To create a reward model for reinforcement learning, we needed to collect comparison data, which consisted of two or more model responses ranked by quality

11 replies

1 recast

3 reactions