Search papers, labs, and topics across Lattice.
Apple
1
0
9
Confident predictions in language models can be surprisingly fragile, and leveraging this fragility can dramatically improve error prediction and uncertainty estimation.
