Scaling Law has been vital to the rapid growth of LLMs over the last few years. Covariant is no stranger to Scaling Law—a concept we’re applying to our Robotics Foundation Model. We launched RFM-1 with the objective of improving the fidelity of the world model by scaling compute, dataset size, and model size. Our latest update demonstrates the Scaling Law in action, with RFM-1 now generating images at 400% higher resolution than previously, increasing the accuracy of AI predictions and reducing hallucinations
See our example images here showing RFM-1 making higher fidelity predictions of what a bin might look like after a specific item is picked.
