In shared autonomy, user input is combined with semi-autonomous control to
achieve a common goal. The goal is often unknown ex-ante, so prior work enables
agents to infer the goal from user input and assist with the task. Such methods
tend to assume some combination of knowledge of the