Search papers, labs, and topics across Lattice.
Department of Computer Science, University College London, The United Kingdom
2
0
4
Hidden misconceptions in student responses can be detected with 84% accuracy, but the cost of false alarms is alarmingly high—8 to 1 against genuine detections.
Small language models can punch above their weight in topic-controlled summarization, rivaling larger models, simply by training on cleverly mixed-and-matched document contexts.
