Conceptredistributing probability mass
Probability Mass Redistribution
A core mechanism in reinforcement learning training of language models. Methods such as policy gradient use reward signals to adjust model parameters, shifting probability mass away from low-reward outputs toward high-reward outputs, rather than creating entirely new capabilities.
Timeline (last 90 days)
Oct 7
RL training redistributes probability mass from ordinary outputs to high-quality outputs through methods such as policy gradients, and does not create entirely new capabilities.
Unverified50%