C. Karen Liu

“Listed as equal advising contributor.”

“The paper introduces RoboToken, a tokenization scheme that converts any articulated robot's embodiment (links, joints, motors), states, and actions into a consistent sequence of continuous-valued vectors.”

“Rather than maintaining separate neural networks for design generation, design evaluation, and robot control, Transformer Transformer consolidates all three into a single diffusion transformer.”

“Co-author with deep expertise in physics-based character animation and contact dynamics. Brings the dynamics modeling and simulation expertise critical to the paper's approach.”

“C. Karen Liu — Stanford University. Prominent researcher in character animation, motion synthesis, and humanoid control. Co-author on OmniH2O and OmniRetarget (cited as [15, 72]), which are foundational to humanoid whole-body control.”

“Mana achieves zero-shot sim-to-real transfer — no real-world demonstration data was collected. The full pipeline generates training data in simulation, trains a visuomotor policy, and deploys directly on a physical Allegro hand.”

“The paper lists Amazon FAR (Fundamental AI Research) as a co-affiliation of all four authors (title page). This is an Amazon-affiliated research output, signaling Amazon's investment in dexterous physical AI capabilities.”

“When real-world demonstrations are difficult to acquire and end-to-end reinforcement learning from scratch is too brittle, we turn to Computer Animation.”