TsinghuaC3I/Awesome-RL-for-LRMs

Comprehensive taxonomy of reinforcement learning techniques applied to reasoning models, organizing papers across reward design (generative, dense, unsupervised), policy optimization algorithms (policy gradient, critic-based, off-policy), and sampling strategies. Covers training resources spanning static corpora (code, math, STEM) and dynamic environments (rule-based, code-based, game-based), plus RL infrastructure and implementation frameworks. Extends to diverse applications including coding agents, browser automation, multimodal understanding, robotics, and scientific tasks with curated paper collections and linked implementations.

2,368 stars.