# RLHF: Reinforcement Learning from Human Feedback — Ouyang 2022 InstructGPT'den ChatGPT'ye

> Source: https://sukruyusufkaya.com/learn/llm-muhendisligi/rlhf-ouyang-2022-instructgpt-chatgpt
> Updated: 2026-08-21T10:08:47.302Z
> Category: LLM Mühendisliği
> Module: Modül 15: RLHF + DPO — Alignment & Preference Optimization
**TLDR:** RLHF'in tam anatomisi: SFT model → reward model training (Bradley-Terry) → PPO RL training. Ouyang 2022 InstructGPT paper, 3-stage pipeline, KL divergence penalty, reward hacking concerns. ChatGPT'nin gizli sosu. Türkçe RLHF zorlukları (human annotator pool, cultural nuances).