Generalist announces GEN-1.5, a one-shot learning model for physical tasks

In a post on X, Generalist announced GEN-1.5, a foundation model for physical tasks. The company claims it can learn and generalize to new tasks from a single 3- to 12-second demonstration without gradient updates. However, there has been no independent verification, and the company provided no task details or benchmarks.

Generalist says GEN-1.5 can learn new physical tasks from a single 3- to 12-second demonstration and respond without gradient updates or fine-tuning. For robot developers seeking to teach robots real-world actions from few examples, the model offers a basis for assessing the conditions under which that learning works.