Overview# Welcome to AReaL’s documentation!# Version History Key Milestones Tutorial Installation Installation (Huawei Ascend NPU) Installation (AMD ROCm) Quickstart Agentic Reinforcement Learning Online RL Training Evaluation Fine-tuning Large MoE Models Archon: PyTorch-Native Training Engine Configurations Code Walkthrough Running GRPO on GSM8K Dataset Best Practices Diagnosing RL Performance Writing Agent Workflows Debugging Guide Handling OOM Issues Performance Profiling Benchmarking FSDP training steps Customization Dataset Data engineering recipes Custom Agent Workflows Custom advantage computation Algorithms Asynchronous RL On-Policy Distillation Direct Preference Optimization (DPO) PPO, GRPO, and Related Algorithms Second-Moment Trust Policy Optimization (M2PO) Proximal Log-Probability Approximation Process Rewards Reference Checkpointing Metrics Tracking Visualize MoE expert loads in W&B Allocation Mode LoRA Reference Megatron-HF Bridge Backend Tree Training RolloutWorkflow Reference Agent Workflow AI-Assisted Development