Title: I built a reward analysis tool for AI alignment — here's why reward hacking is harder to detect than you think

When you train an AI with reinforcement learning, the reward function is supposed to guide it toward the behavior you want. But what happens when the model finds ways to maximize reward without actually doing what you intended? Analyzes reward signal distribution across training episodes If you're i