Masoud's Notes

Part 6: Cutting Out the Middleman – Policy Gradient Methods

6 min read   ·   April 08, 2026

2026   ·   reinforcement-learning   machine-learning   ai   policy-gradients   actor-critic   ·   reinforcement-learning

Part 7: The Complete Alignment Pipeline — From SFT to Advanced RL

A complete guide to how PPO, DPO, and GRPO transform language models from pattern copiers into reasoning agents.

8 min read   ·   April 08, 2026

2026   ·   reinforcement-learning   machine-learning   ai   ppo   dpo   grpo   ·   reinforcement-learning

Part 5: Breaking the Table – Function Approximation

6 min read   ·   April 07, 2026

2026   ·   reinforcement-learning   machine-learning   ai   deep-rl   ·   reinforcement-learning

Part 4: A Cliffhanger - Comparing SARSA, Q-Learning, and Expected SARSA

7 min read   ·   April 06, 2026

2026   ·   reinforcement-learning   machine-learning   ai   q-learning   sarsa   ·   reinforcement-learning

10 RL Stability Tests That Detect Collapse Before Rewards Drop

How to catch reinforcement learning collapse early with the right stability tests, before headline reward metrics tell you everything is fine.

9 min read   ·   April 06, 2026

2026   ·   reinforcement-learning   mlops   deep-learning   ai   ·   ai-research