There are typos in the document. It says re-encoding the base pose doesn't give the base pose, when the paper says it does. I think I meant that they don't show re-encoding an arbitrary input pose.
The videos are missing from the presentation, they can be found
- Semantic Deformation Transfer
- Deformation Transfer for Triangle Meshes
- Linear Rotation Invariant Coordinates ( missing link, or shoddy acm streaming )
and here: