In-Context Iterative Policy Improvement for Dynamic Manipulation

Mark Van der Merwe, Devesh Jha

Attention-based architectures trained on internet-scale language data have demonstrated state of the art reasoning ability for various language-based tasks, such as logic problems…