Skip to content

Reinforcement learning

Reinforcement learning is a class of of ML that uses dynamic feedback from environments to enable more successful outcomes.

MANAGEN TODO: Insert general description in here.

In the context of Generative AI, it may be considered that each token that is generated is an action taken in the state-space of tokens. Consequenetly, RL has been used as a method for improving, other Geerative models using feedback methods.

Notable research

Learning to Model the World with Language Uses multimodal agents to build world models to act in.

Also introduces Homegrid evaluation game. Fun continuous multimodal agent. Github image