David A. Clifton

The paper introduces MedMisBench, a benchmark to measure LLMs' ability to maintain correct medical judgment under misleading context, showing significant drops in accuracy when exposed to adversarial inputs.

This paper develops an AI-assisted in-silico trial to optimize osmotherapy protocols for ischaemic stroke by predicting intracranial pressure and personalizing treatment based on patient-specific data.

This study quantifies antibiotic resistance in tuberculosis using genomic data and MIC measurements to identify genetic determinants of resistance and hypersensitivity across multiple drugs.

ProSelfLC improves DNN robustness by progressively trusting low-entropy predictions and correcting labels during training.

This paper uses machine learning on whole genome data to improve source attribution of Campylobacter jejuni infections, achieving higher accuracy than traditional methods.

GenomegaMap estimates genome-wide dN/dS ratios across 10,209 tuberculosis genomes to identify selection pressures shaping the pathogen's evolution.