unreviewed draftNo one has checked this end to end. Treat the math, code, and citations as unverified. Source on GitHub.
Topic: GRPO and RL with verifiable rewards. Goes with: lecture L12, L15. Status: planned. To be built next.
The plan: derive the GRPO advantage and show why a value network isn’t needed; sketch the R1-Zero training loop; do the GRPO exercise (links to exercises/15-grpo-rlvr/).