Covariant Robotic Foundation Models Quiz

Answers and explanations appear as you take the quiz.

1. What was the name of Covariant's foundation model introduced in 2024?
   • RFM-1
   • GPT-4
   • Helix
   • RT-X
2. What did Amazon’s 2024 commercial agreement with Covariant include?
   • A purchase of every Covariant customer contract
   • A non-exclusive license to Covariant’s models and hiring several founders and employees
   • A ban on Covariant serving other customers
   • An agreement to manufacture consumer phones
3. What core idea does RFM-1 borrow from large language models?
   • Storing a fixed lookup table of every object
   • Using only handwritten rules per item
   • Predicting the next token across multiple data types to generalize
   • Rendering 3D animations of grasps
4. What was RFM-1 designed to help warehouse robots do?
   • Replace all warehouse workers
   • Drive passenger cars on highways
   • Render 3D animation frames
   • Handle picking tasks involving varied or unfamiliar items
5. Who were among Covariant's notable founders?
   • Pieter Abbeel and Peter Chen
   • Brett Adcock alone
   • Marc Raibert
   • Demis Hassabis
6. What does it mean that RFM-1 is 'multimodal'?
   • It can only process text
   • It learns from several data types like text, images, video, and robot actions
   • It works on multiple robots but only one task
   • It uses multiple separate models glued together at runtime
7. What physical prediction capability did Covariant describe for RFM-1?
   • Translating warehouse signs into music
   • Measuring human speech quality
   • Generating video predictions of how objects may respond to robot actions
   • Rendering a scene from a NeRF
8. How many parameters did Covariant report for RFM-1?
   • Eight million
   • Eight trillion
   • Eighty thousand
   • Eight billion