🧭 Roadmap
›
08 · Advanced Alignment
›
Reinforcement Learning (RLHF)
🌙
☀️
08 · Advanced Alignment
Reinforcement Learning (RLHF)
Shaping model behavior with human preference signals
🔴 MUST
🧠 Understand
👁️ Visualize
✏️ Calculate
⚖️ Compare
🗣️ Explain
10 subtopics — click to study
Why RL for LLMs?
🔴
Reward Model
🔴
Human Preference Data
🔴
PPO (Proximal Policy Optimization)
🔴
Policy Model
🔴
KL Divergence Constraint
🟡
DPO (Direct Preference Optimization)
🟡
RLHF Pipeline
🟡
Alignment Goal
🔴
GRPO / REINFORCE variants
🟢
← Back to Roadmap
This page
0:00
This session
0:00
Total (all pages)
0:00
Today
0:00
▶ Running
📊 View all study stats →
0:00
studying