thinkwee/AgentsMeetRL

Curated taxonomy of 205+ open-source repositories combining LLM agents with reinforcement learning, organized across 16 categories from base frameworks (veRL, OpenRLHF) to specialized domains like embodied agents and safety alignment. Emphasizes technical choices in RL algorithms, reward models (external verifiers, rule-based, model-based), and training environments, with an interactive dashboard for exploration. LLM-assisted code analysis identifies projects with multi-turn interactions or tool use, though manually reviewed for accuracy.

853 stars. Actively maintained with 6 commits in the last 30 days.