Installing a bundle creates a private copy you can edit. The published bundle stays unchanged.
109 bundles
Shard training across multiple GPUs/nodes with proper gradient synchronization
Orchestrate parallel agent rollouts across environments at scale
Self-critique and revision loops using model-generated feedback
End-to-end: collect preferences → train reward model → optimize policy
When to use online rollouts vs. offline datasets, and how to blend
Best-of-N sampling with a reward model; simplest "RL" that actually works
Classic REINFORCE with variance reduction, the foundation of policy gradient methods
Build Group Relative Policy Optimization as used in DeepSeek-R1
Build Direct Preference Optimization, understand when it outperforms PPO
Build Proximal Policy Optimization from scratch, understand clipping and advantage estimation
Convert real production failures into new eval cases automatically
Run the same eval suite across Haiku/Sonnet/Opus (or GPT-4/Claude/Gemini) and compare