GPG (Group Policy Gradient)

Minimalist RLVR method that strips away surrogate ratio clipping, reference-model KL divergence, and critic networks altogether, directly optimizing the group-normalized REINFORCE objective with dynamic thresholding of zero-variance prompts.

Useful when memory constraints forbid keeping $\pi_{\text{ref}}$ and $\pi_{\theta_{\text{old}}}$ alongside $\pi_\theta$.