TF-Agents
TensorFlow-based framework focused on parallel RL algorithms, differing from verl's specialization in post-training RL for LLMs.
RLtools
A C++ header-only deep RL library, more general-purpose compared to verl's LLM-specific focus.
ReAgent by Meta
Decision-based RL framework with a broader scope beyond LLM post-training.
OpenSpiel by DeepMind
Provides environments and algorithms for general RL research, not specialized for LLM post-training.
Amazon SageMaker RL
Managed RL service with simulation integration, differing from verl's open-source framework for LLMs.