Core
RLHF (Alignment)
RLHF aligns a model to human preferences in three stages: supervised fine-tuning on demonstrations, training a reward model from human comparisons of outputs, then optimizing the policy with RL against that reward while a KL penalty anchors it to the base model. It shapes behavior and tone rather than facts, and FDE loops probe it because reward hacking and the KL anchor are where deployments actually go wrong.
a free account unlocks the core curriculum tier · no card
Related concepts
Practise this in questions
LLM & GenAI FundamentalsWalk through the RLHF pipeline (SFT, reward model, PPO) and explain how DPO simplifies it. →hard · PremiumAI Security, Privacy & GovernanceReproduce-from-paper: design a production-safe Constitutional-AI-style fine-tuning pipeline that aligns a chatbot to a set of rules →hard · PremiumLLM & GenAI FundamentalsGRPO vs PPO-based RLHF: how does group-relative reward let you drop the value model? →hard · PremiumLLM & GenAI FundamentalsTrain a reward model for a coding agent from 100K noisy human scores biased toward short solutions. →expert · PremiumLLM & GenAI FundamentalsWhat's the difference between a base model and an instruct model, and why would a customer care? →easy · FreeAI Security, Privacy & GovernanceWhat is data/model poisoning, and how would you detect a backdoor introduced through fine-tuning or embedding data? →medium · Free