Search papers, labs, and topics across Lattice.
1
0
4
3
Task progress in vision-language-action models can be read directly from their internal representations, even before task-specific training, revealing a surprising depth of interpretability.
