Cosmin: RL for LLMs review - part 1

# Cosmin: RL for LLMs review - part 1

This Tuesday we review fundamentals of RL for LLMs, going through policy gradient, reinforce and grpo and recent GRPO followups.

We'll cover material from Nathan Lambert's RL course lectures 3 and 4 (math and practical/coding).

Preparation:
Cosmin will do a presentation. Skim the material below so we can have a productive q&a.

Resources:
Cosmin's cheat sheet on the evolution of RL for LLM algorithms
https://claude.ai/code/artifact/2b8a950c-6006-4bad-a3df-9154f5db5166

Interconnects RL Course - lecture 3 and 4
https://rlhfbook.com/course
math - https://rlhfbook.com/teach/course/lec3-chap6-p1/#1
practical/coding - https://rlhfbook.com/teach/course/lec4-chap6-p2/#1