Pearl

ART (Agent Reinforcement Trainer) trains multi-step agents for real-world tasks using GRPO reinforcement learning, enabling on-the-job training for models like Qwen, Llama, and more.

OpenRLHF is a high-performance agentic RL framework based on Ray and vLLM, offering PPO, DAPO, and REINFORCE++ algorithms for large-scale training of agents and vision-language models.

Hive is a production-ready multi-agent execution harness providing state management, failure recovery, observability, and human-in-the-loop control with auto-generated multi-agent topologies for complex business workflows.

A modular high-level library from Meta to train embodied AI agents across a variety of tasks and environments.