Awesome-RL-for-LRMs

Why we included this project

Keeping up with reinforcement learning for large reasoning models means reading a lot of papers, and this survey does the sorting for you. It organizes the literature around the decisions you actually face when training a reasoning model: how to design rewards, which policy optimization method to use, how to sample, and what training resources are out there. The paper list is grouped by concrete distinctions like generative versus dense rewards, critic-based versus critic-free algorithms, and rule-based versus code-based environments, so you can find the work closest to your own setup without wading through everything. It also points to the frontier models and the infrastructure used to train them, which helps when you are weighing which open-source recipes to try. The maintainers keep adding new work and take issue submissions for anything missing, so the list stays current as the field shifts.