Search papers, labs, and topics across Lattice.
1
0
3
Forget finetuning encoders: representing human motion as structured text unlocks surprisingly strong performance on motion understanding tasks by directly leveraging LLMs' pretrained knowledge.
