Didactic knowledge or Clinical Cases? How Data Types Shape Medical Large Language Models

Recent research explores how data composition affects Medical Large Language Models. Experiments show clinical data enhances clinic-oriented tasks while remaining competitive in knowledge-intensive areas, unlike didactic data. Separately, a new study provides a probabilistic perspective on Large Language Models. It integrates tools by describing LLMs through probability measures on token sequences. This framework formulates training as a maximum-likelihood estimation problem using stochastic gradient methods and views text generation as the sequential simulation of a stochastic process.