Mayank Kejriwal

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."

Despite significant AI investment by NIH, a concerning gap exists between research and real-world clinical deployment, coupled with a stark underrepresentation of AI applications addressing health disparities.

LLMs can bootstrap scalable aspect-based sentiment analysis by identifying key aspects, which can then be used to train efficient classifiers on massive datasets.