Self-Distilled Policy Gradient Unifies RL and Distillation at Token Level

Users find the Self-Distilled Policy Gradient method fascinating because it unifies RL and distillation via token rewards.

1 comment with sentiment. View comments.