Search papers, labs, and topics across Lattice.
University of Southern California Los Angeles
1
0
3
Small language models can reliably signal uncertainty without any parameter updates, simply by observing how consistently they answer when given the option to say "I don't know."
