#Stanford

'OpenTSLM integrates time series as a native modality in pretrained LLMs to interpret ECGs, EEGs, and wearable data, achieving major performance and scalability gains over VLM and text-based approaches'

'Stanford released MedAgentBench, the first large-scale FHIR-compliant benchmark that tests LLM agents in realistic EHR workflows, revealing strong retrieval skills but gaps in safe multi-step action execution.'