Search papers, labs, and topics across Lattice.
1
0
3
Small language models can achieve surprisingly high accuracy (84.3%) in extracting structured information from specialized clinical text, even outperforming larger, off-the-shelf models.