Tag
This talk by Will Brown of Primordial AI discusses techniques for scaling Reinforcement Learning to complex, real-world tasks where rewards are not verifiable, using methods like anchoring, LLM judges, and simulation.